TruthIsAll: Plain English—“The Past is Prologue"- 1st Line Fraud Defense

[autorank]

Happy New Year DU. This IS the year we start to roll in a big way. I’ve done the election news thread for months and in the past 40-60 days, I’ve noticed a steady increase in MSM press covering stories. It’s started with small town papers a while back, not it’s moving to suburban and mid size city dailies. We’re not far from real coverage.

Here’s the much requested TruthIsAll in plain English, no Excel needed (unless you want it). The past is prologue. Election Integrity Movements all over the world use exit polls as their FIRST LINE DEFENSE against election fraud. Here’s how to understand the nonsense and also protect our democracy. We need good candidates, hard work (especially GOTV), and vigilance against election fraud. Here are the basics for exit polls.

The numbers say Kerry won. Easily. It's all here.
Want to see how the exit polls matched the pre-election polls?
UNDERSTANDING IS THE BEST PROTECTION AGAINST RECURRENCE.
----------------------------------------------------------------------------------
TIA Pre-election Popular Vote and EV projections:

Final Nov.1, 2004 Election Model
Click the links to view the numerical tables and corresponding graphs.
Along with a full, clear explanation of the methodologies.

http://www.geocities.com/electionmodel/

TIA Post-election analysis:

Polling Simulation Model (HTML): overview, model description and graphs
http://www.geocities.com/electionmodel/TruthIsAllPollingSimulation.mht

Polling Simulation Model (Word DOC): overview, NUMERIC TABLES and graphs:
http://www.geocities.com/electionmodel/MonteCarloPollingSimulation.doc

Interactive Election Model (Word DOC): Comprehensive election analysis
http://us.share.geocities.com/electionmodel/InteractiveElectionSimulationModel.doc

IF YOU HAVE EXCEL, TRY THESE...

Polling Simulation Model (Introductory):
http://us.share.geocities.com/electionmodel/MonteCarloPollingSimulation.xls

Interactive Election Simulation Model (Comprehensive):
http://us.share.geocities.com/electionmodel/InteractiveElectionSimulation.xls

PRESS the F9 key to CALC.
THAT'S IT.

________________________________________________________

A New Years challenge to all who still believe Bush won:
Find ONE plausible Bush win scenario.

1-Download the Interactive Election Model
2-Select the "NatExit" sheet.

Play "what-if" using the National Exit Poll "How Voted in 2000" demographic.
Try various 2004 vote shares for Kerry, Bush and Other for those who
a) did not vote
b) voted for Gore
c) voted for Bush
d) voted for Nader

The only restriction is the WEIGHTING constraint:
The MAXIMUM Bush 2000 voter percentage (weight) of the 2004 total vote
was 39.8% (48.7/122.3 mm). See why below.

Hint:
You can save a lot of work if you first view the "Voted2000" sheet.
The "Voted2000" sheet displays a comprehensive SENSITIVITY ANALYSIS
of Kerry's national vote and margin to the following parameters:
1) Gore voter turnout (assume 100% Bush 2000 voter turnout).
2) Kerry's percentage of those who DID NOT VOTE in 2000.
This analysis was originally presented in the "Clincher" DU thread.

According to the National Exit Poll timeline, Kerry's share
of those who did not vote in 2000 was:
59% at 7:33pm (11027 respondents)
57% at 12:22am (13047)
54% at 1:25pm (13660 Final)

Here's proof that the MAXIMUM Bush 2000 voter percentage (weight)
of the 2004 total vote was 39.8% (48.7/122.3 mm).

1- Bush got 50.456 million votes in 2000.
2- About 1.75 mm died prior to the 2004 election.
3- Therefore, AT MOST 48.7 mm could have voted in 2004.
4- The 1:25pm Final NEP 43% weighting assumes 52.6 mm Bush 2000 voters.
This is obviously a physical and mathematical IMPOSSIBILITY.

If you think you found a feasible Bush winning scenario, post it on:
progressiveindependent.com or democraticundergound.com

[EFerrari]

1. I've notice the up tick in coverage, too. We're about to

reach critical mass. Let's keep rolling.

:kick:

[autorank]

3. This is a nice piece for the press, quick study.

I see stories that I know are "activist informed" from local media. The best example is the GuvWurld coverage in Humbolt County CA for his toss Diebold out campaign. But there's more.

The main point to remember in the coming months is: BUSH IS NOT THE LEGITIMATE PRESIDENT.

HE LOST THE ELECTION. OPEN YOUR EYES. LOOK AT THESE PROOFS, THE CONYERS REPORT, THEH FRITAKIS/WASSERMAN BOOK, ETC. ETC. It all fits together.

...an of course, he absolutely lost in 2000. Impeach all of the 5 who sElected him, every one of them, voice vote!

[EFerrari]

6. Guv's work has been outstanding.

It might not be a bad idea for us to check our local media outlets and see if/how they've shifted their positions and keep helping them along.

The SF Chronicle is still entrenched in its anti- Kevin Shelley position and may not budge for a while. But another local weekly here most likely will. I'm gonna go look at their coverage and see where they are on the continuum.

:kick:

[liam_laddie]

9. recent NYT ink

In the 12-25-05 Sunday NYT, (5) letters to the editor were
published on the editorial page regarding elections' systems
honesty; this indicates that some editors are aware.
Then in the 1-1-05 issue, Business section page 2, James
Fallows (Atlantic Monthly correspondent) discusses technology
issues, and ends his piece with this paragraph, copied below the
link. I don't like his reference to the Carter-Baker Shamission, but
the link to VerifiedVoting can't hurt.

<http://www.nytimes.com/2006/01/01/business/yourmoney/01techno.html>
-----------------------------------
A final point, and no joke: The most important tech-related question of 2006 is
whether America's electoral system will become more trustworthy, or less, as it
becomes more computerized. The technical steps needed to make computerized
voting fair, accurate and accountable have been spelled out repeatedly (and
discussed here), most recently in a report by former President Jimmy Carter and
James A. Baker III, the former secretary of state. Many states have passed
appropriate laws, but many other states and the federal government have not.
If anyone is looking for a political cause, I suggest this one. <VerifiedVoting.org>
is a good place to start.
-----------------------------

[EFerrari]

15. Excellent!

:kick:

[paineinthearse]

2. Thanks for your perserverence

"The numbers say Kerry won" Too bad he didn't believe in himself in the wee hours of November 3.

Cultivating personal relationships with local media is the key. It's starting to affect the big fish, witness the recent development involving the NYT's public editor.

KEEP ON!

[autorank]

4. Thanks for your perseverence...Conyers inspsired this whole movement.

He knows the deal, and knew election night.

Let's run them out of town.

[paineinthearse]

5. Let's run them out of town.

but...but...where's your party unity????? Can we say that?

[autorank]

11. Oh, I'm sorry, "Let's, uh...let's have a committe study if we should ...

:sarcasm:... consider, given all the more prudent alternatives, running them out of town (only if it doesn't hurt our move to the middle and well, only if it doesn't set us up to be run out of town; but let's let the committee meet, Donna will handle that).

[AuntiBush]

22. He Should Have Been Nu 1 on that 2005 Award (Conyers)

Would luv to lead the "run them out of town" pack! :) Even if I have to crawl...

[understandinglife]

7. Great start to the year we restore America and send a bunch of ...

... folk to jail!


Peace.

[autorank]

8. "Socialism for the rich, free enterprise for the poor."

Well, the way things are going, we'll soon be "the wretched of the earth." It's now two in a row, stolen, 2000 & 2004. If you're a white collar crook, you get special justice. To wit, 17 of 34 white collar convictions resulted in "deferred judgments" -- a method used primarily for 1st time juvenile offenders. You're guilty and you admit it. Then 6-12 months later, if you're "clean" your charges are dropped.

It's time for a real acounting...of votes and those entrusted to handle them. Unless we press this, as citizens, people will say of us, "We the people have no clothes...

[mom cat]

10. Thanks for the numbers...they validate what I know in my gut.

[Kurovski]

12. K&R.

And a happy new year to you, autorank.

[Stevepol]

13. K&R!!

[tn-guy]

14. I know this is futile but ..........

Let's try to review just one more time. My comments in italics.

1. "The 1:25pm Final NEP 43% weighting assumes 52.6 mm Bush 2000 voters." (This is because that is what was reflected in the answers received in the sample.)

2. "This is obviously a physical and mathematical IMPOSSIBILITY." (And therefore demonstrates that the NEP is unreliable; thus, any further analysis will simply compound error upon error.)

I must say that this whole discussion by TIA and Autorank is beginning to resemble the sort of reasoning the ancient church fathers used to demonstrate that Galileo had to be wrong. In both cases the argument proceeds from faulty assumptions in an attempt to prove a preconceived notion that is based on an unshakable faith. Their dedication, perseverance and diligence are admirable but, unfortunately the only useful result is to show that the NEP is unreliable; something that pretty much everyone already knew.

[Cocoa]

19. this post makes you a heretic

known nowadays as "naysayer." :-)

[Name removed]

23. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Usrename]

24. Are you really not able to trust math?

All TIA is doing is providing an arithmetic proof that the ballots were not counted correctly. Has nothing to do with garbage. Unless you are going to argue that the gender info from the NEP is garbage too. That thousands of Bush voters dressed in drag to deceive the exit pollsters.

Wake up and smell the numbers.

[autorank]

27. The Bush "drag vote" is greatly overestimated;)

Great points Usrename. The facts are very threatening. Why else would math based posts draw such immediate and spirited, not to mention, misleading and disingenuous responses?

Happy New Year and I promise to stop calling you "Username" - U.s.r.e.n.a.m.e ...!

[tn-guy]

30. sorry

It appears I am not permitted to doubt the prevailing wisdom.

[rman]

31. you are permitted to doubt anything you want,

but your doubt is probably not going to convince many people of anything of any significance.

[Cessna Invesco Palin]

61. It must be a slow day in the election asylum forum.

I have been following this "debate" since before the 2004 election, and at this point I believe that there's nothing you can say or do to make these folks see logic or reason. It is now nothing more than an ego-driven attempt to explain away the monumentally wrong (even taking into account supposed election fraud) conclusions drawn by TIA in the run-up to the 2004 election. In all this time, the only thing this debate has taught me is that sales of Microsoft Excel should be regulated in the same manner as handguns.

[foo_bar]

67. related reading

"What interests me most in this book, other than Dudley's enthusiastic and deft writing, are his catalogues of crank behavior -- for instance, how they can go from enthusiastic amateur to demented conspiracy theorist rather than simply admit nobody's paying attention to them because they're wrong."
http://www.amazon.com/gp/product/0883855070/002-1759993-6960045?v=glance&n=283155

"Unfortunately, some cross that ill-defined line and refuse to consider the evidence contradicting their claims. Many fail to
understand that mathematical truth is incommensurate with physical theory."
http://www.amazon.com/gp/product/0883855070/002-1759993-6960045?v=glance&n=283155

"The true believer is sometimes referred to as a crank, and Underwood Dudley has written an illuminating book, "Mathematical Cranks", which was published by The Mathematical Association of America. People who are described in that book also demonstrate the same mentality of those who started the cults described by Gardner. All have little regard for facts, are egomaniacal in the extreme and consider all of their critics to be insignificant, misguided and perhaps mentally deranged."
http://www.amazon.com/gp/product/customer-reviews/0486203948/ref=cm_rev_next/002-1759993-6960045?%5Fencoding=UTF8&customer-reviews.sort%5Fby=-SubmissionDate&n=283155&s=books&customer-reviews.start=11

"A mathematical proof of the impossibility of doing something does not deter some people from claiming that they have done it. Dudley labels such people as cranks, and he describes many of them in this book. While the reading is entertaining, it is also somewhat depressing, in that no amount of logic or reason can convince a crank that their work is flawed."
http://www.amazon.com/gp/product/0883855070/002-1759993-6960045?v=glance&n=283155

P.S. "All of us are gullible to one degree or another-- the secret is to puncture unmerited beliefs as much as possible, before they grow into full-scale dogmas that one ends up hauling around like bags of bricks."
http://www.amazon.com/gp/product/customer-reviews/0486203948/ref=cm_rev_next/002-1759993-6960045?%5Fencoding=UTF8&customer-reviews.sort%5Fby=-SubmissionDate&n=283155&s=books&customer-reviews.start=11

[Name removed]

26. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Name removed]

28. Deleted message

Message removed by moderator. Click here to review the message board rules.

[autorank]

29. Validity -- Utility -- International effect...of a positive kind.

The 1:25 p.m. Nov. 3 Final exit poll (13660 respondents) which Bush won by 51-48% is simply not possible, not unreliable...impossible..

Does "everyone" also know that the Final was matched to the vote count 13 hours after the election eve final. Even with all that time, they couldn't cook the books properly and invalidate the election eve/early a.m. exits, which which are real polls.

Since the the Final 13660 Exit Poll is bogus, what does it say about the vote count?

The Final Exit Poll (13660 respondents) is the only one which Bush "won".

Kerry won the 12:22am time line of 13047 respondents by the same 51-48%;; and that was with a 41/39% Bush/Gore split, also not possible. Imagine if plausible weights were used...the final poll would have shown Kerry an even bigger winner.

Exit polls are used all over the world to catch election fraud, often endorsed by people like Carter, Lugar and others. They were used with great success in the "Orange Revolution" as the trip wire to challenge elections, again with broad international endorsement. Why" Because there are many out there who steal elections. In the Ukraine, the election thieves were also poisoners--they tried to kill the challenger to Russian hegemony. If people poison others or start huge wars based on lies in which hundreds of thousands die and are injured; and if these people exist in a culture with a history of election fraud; then it's entirely reasonable to assume that these same people, East or West, would have no qualms about stealing elections. Any available tools are appropriate. This is one of them.

[Dongfang Hong]

16. So we've proven the exit poll was unreliable.

Great. Certainly this is enough to cover the massive burden of proof a claim like a stolen election would require.

[autorank]

17. Wrong, the polls prove the election was stolen.

The head of the polling company Edison Mitofsky was not a Bush Campaign director as Ohio Secretary of state Blackwell was in 2004--he headed the Bush campaign and also ran the voting.

The head of the polling company, Edison Mitofsky did not write and sign a letter saying he was committed to doing anything necessary to get Bush elected. Wally O'Dell, head of Diebold said just this to a fund raising group of stalwart republicans called "The Rangers."

The exit poll interviewers for Edison Mitofsky did not make people stand in line for hours to get interviewed; did not threaten to challenge their credentials to be interviewed; did not move locations on them at the last minute. Ohio Secretary of State Blackwell did all of these things as did many county election officials in Ohio and elsewhere.

The Edison Mitofsky company was able to produce a paper trail of all of their interviews and data activities. The Secretary of State of Ohio, Florida, and every other state which used an electronic touch screen and/or electronic tabulator cannot produce any paper trail or even an audit trail.

The polls on election day in 2004 were more reliable, more honest, and much less biased than those who administered the voting and the polling companies kept their records, which was not possible for the machine users.

Of course, after a final prediction around 12:30am 11/03/04 that Kerry won by 3% (based on 13,000 interviews, Margin of Error, 1%), the head of th polling company had a big problem. He solved that 12 hours later by revising his polls to take into account the final vote count. This produced a fifth national poll, nothing more than a revision, which showed a * victor.

[Dongfang Hong]

35. The polls suggest that a theft was possible.

But suggestion is not proof. The path of the Kennedy assassination bullet suggests multiple shots from multiple angles. But yet recent ballistic tests have proven that it is possible, though unlikely, to make the 'magic bullet' shot.

Similarly, the best counterargument to "but it was very unlikely" is that unlikely things do happen.

[autorank]

36. Why not ask for an investigation? Or is that asking too much?

Of course it's not. This isn't a court of law. People can argue and think what theywant. I, for example, am convinced OJ is guilty. Yes he was let off. Am I supposed to change my mind because of some court based on courtroom standards? I take it you mean that "it was very unlikely" that Bush won. Good point. I'm convinced he lost. In any event, there should be a thorough, independent investigation, hard hitting, people under oath, etc.

[Yupster]

99. Yes and the 1988 exit poll

showed Dukakis and Bush tied though Bush won the election by eight points.

And the 2000 exit poll in Alabama showed Gore nudging Bush by 1 % though Bush won Alabama by 15 %.

And in every election since 1988 when the exit polls were first published the Democratic candidate scored higher than he did in the actual election every single time.

Which only goes to prove that the exit polls are never wrong.

[OnTheOtherHand]

103. ooh, "naysayer" company! the 1988 facts are a bit fuzzy

I believe there were four exit polls in 1988. I've looked at the data from CBS and ABC. The CBS raw data do show Dukakis and Bush practically tied, but that really is misleading, because there pretty clearly are too many black respondents in the sample. If I had to guess, that probably was not due to higher response rates among blacks; it probably had to do with the precinct sample, and actually may have been deliberate. Regardless, CBS probably could adjust for it on the fly. (Still, CBS did call Illinois wrong that year -- I know they didn't call it from exit poll data alone.) So, I've come to think that the 1988 CBS exit poll isn't the best proof that the exit polls can be wrong; rather, it's the best proof that plugging raw percentages into a margin-of-error formula is wrong. Not that TIA is likely to make that distinction, but the reality-based can try.

Ironically, the 1988 ABC raw data appear to give Bush a comfortable lead over Dukakis, yet Richard Morin has written that the poll showed the election to be a "dead heat." It's perfectly possible, for reasons that are hard to explain in a DU post.

It's not quite right that the exit polls were "first published" in 1988. But it's certainly right that the estimates have favored the Dems on net in all five of those elections, '88-'04. The Alabama 2000 example is good, because I don't think many observers think that Gore actually won Alabama that year. Personally, I think New York '04 is also a good example: did Kerry really win by approximately 31.3 points? I think not.

[bleever]

18. 2006 will be a watershed year for the truth.

And...I'm still waiting for someone to prove * won the election!

Why don't they just PROVE IT. Use math. Just come up with a plausible SCENARIO by which he won with the numbers we have.

Is that so much to ask?

[AuntiBush]

20. K&N for Others To See...

This should be repeated daily from here on out here.

I commend you AutoRank! Super :thumbsup: for all your dedicated, hard work!

[AuntiBush]

21. Passing Your Web Links Along to Everyone I know Out Here, too!

[autorank]

25. Great. The "Word Doc" files are great --"Word" yes, simple, easy., etc.

[autorank]

32. Based on 2000, Kerry needed just 85% of Gore voters: HE WON--TIA

From TruthIsAll:
												
		THE NATIONAL EXIT POLL 
           HOW VOTED IN 2000 DEMOGRAPHIC								
												
We will show that National Exit Poll Bush 2000 weights are
mathematically
IMPOSSIBLE in both the preliminary and final timelines.

Time Respondents Bush/Gore Winner
12:22am 13047  (41%/39%): Kerry 51-48%
 1:25pm 13660  (43%/37%): Bush  51-48%

In 2000:								
Bush had 50.456 mm votes, 41.30% of the 2004 total (122.17).
Gore had 51.0 mm votes, or 41.75%.
Nader had 2.883 mm votes (2.36%).																			
The annual U.S. death rate is 0.87%, or 3.5% over four years.
Let's calculate revised MAXIMUM weightings, subtracting voters
who
died:									
							Max weight				
		       Votes Pct 2004	Died	Alive	Pct 2004				
		Gore 	51.00	41.75%	1.79	49.22	40.28%				
		Bush 	50.46	41.30%	1.77	48.69	39.85%				
		Nader	2.88	2.36%	 0.10	2.78	2.28%				
												
Using the maximum weights and assuming that ALL 2000 voters
turned 
out in 2004, Kerry is the clear winner in BOTH
timelines.														
	 
VOTED IN 2000

                          12:22am		 1:25pm
Voted   Weight Votes	Kerry	Bush	Nader	 Kerry	Bush	Nader

No	17.58%	21.48	57%	41%	2%	54.%	45%	1%	
Gore	40.28%	49.22	91%	8%	1%	90%	10%	0%	
Bush	39.85%	48.69	10%	90%	0%	9%	91%	0%	
Other	2.28%	2.79	71%	21%	8%	71%	21%	8%	
			
Total 	100%		52.28%	46.78%	0.94%	50.96%	48.69%	0.36%	
	Vote 	122.17	63.88	57.15	1.14	62.25	59.48	0.44	
		Kerry margin: 6.73mm	Kerry margin: 2.77mm		
												
For the sensitivity analysis, we assume 
a)12:22am NEP timeline
b)100% Bush 2000 voter turnout 
c)Declining Gore voter turnout from 100% to 71%. 

The analysis shows that even with a clear Bush turnout
advantage, 
there is not one plausible scenario of a Bush victory.

At the 12:22am timeline:
Kerry won 57% of those who DID NOT VOTE in 2000.

For Kerry to TIE Bush, he needed just 80% Gore voter turnout.
Very plausible.

For Bush to WIN by 3 mm votes, he needed a maximum 71% Gore
voter turnout.
Not plausible.

At the 1:25pm Final timeline:
Kerry won 54% of those who DID NOT VOTE IN 2000

For Kerry to TIE Bush, he needed 85% Gore voter turnout. 
Very plausible.

For Bush to WIN by 3 mm votes, he needed a maximum 77% Gore
voter turnout.
Not plausible.

												
		National Exit Poll Sensitivity Analysis 				
                
                New Voter Share and Gore 2000 Voter Turnout
                Impact on Kerry Vote Percentage and Margin
	         (assume 100% Bush 2000 voter turnout) 								
													
		       Timeline        7:33pm		12:22am			1:25pm								

	Gore	Revised	Did Not				Kerry Percentage of New Voters						
	Turnout	Weight	Vote	60%	59%	58%	57%	56%	55%	54%	53%	52%	51%
													
							Kerry National Vote 						
													
	100%	40.28%	17.58%	52.81%	52.64%	52.46%	52.28%	52.11%	51.93%	51.76%	51.58%	51.41%	51.23%
	99%	39.88%	17.98%	52.69%	52.51%	52.33%	52.15%	51.97%	51.79%	51.61%	51.43%	51.25%	51.07%
	98%	39.48%	18.39%	52.56%	52.38%	52.19%	52.01%	51.83%	51.64%	51.46%	51.27%	51.09%	50.91%
	97%	39.08%	18.79%	52.44%	52.25%	52.06%	51.87%	51.69%	51.50%	51.31%	51.12%	50.93%	50.75%
	96%	38.67%	19.19%	52.31%	52.12%	51.93%	51.74%	51.54%	51.35%	51.16%	50.97%	50.78%	50.58%
													
	95%	38.27%	19.60%	52.19%	51.99%	51.80%	51.60%	51.40%	51.21%	51.01%	50.82%	50.62%	50.42%
	94%	37.87%	20.00%	52.06%	51.86%	51.66%	51.46%	51.26%	51.06%	50.86%	50.66%	50.46%	50.26%
	93%	37.46%	20.40%	51.94%	51.73%	51.53%	51.33%	51.12%	50.92%	50.71%	50.51%	50.31%	50.10%
	92%	37.06%	20.80%	51.81%	51.60%	51.40%	51.19%	50.98%	50.77%	50.56%	50.36%	50.15%	49.94%
	91%	36.66%	21.21%	51.69%	51.48%	51.26%	51.05%	50.84%	50.63%	50.42%	50.20%	49.99%	49.78%
													
	90%	36.26%	21.61%	51.56%	51.35%	51.13%	50.91%	50.70%	50.48%	50.27%	50.05%	49.83%	49.62%
	89%	35.85%	22.01%	51.44%	51.22%	51.00%	50.78%	50.56%	50.34%	50.12%	49.90%	49.68%	49.46%
	88%	35.45%	22.42%	51.31%	51.09%	50.86%	50.64%	50.42%	50.19%	49.97%	49.74%	49.52%	49.30%
	87%	35.05%	22.82%	51.19%	50.96%	50.73%	50.50%	50.28%	50.05%	49.82%	49.59%	49.36%	49.13%
	86%	34.64%	23.22%	51.06%	50.83%	50.60%	50.37%	50.13%	49.90%	49.67%	49.44%	49.21%	48.97%
													
	85%	34.24%	23.62%	50.94%	50.70%	50.47%	50.23%	49.99%	49.76%	49.52%	49.28%	49.05%	48.81%
	84%	33.84%	24.03%	50.81%	50.57%	50.33%	50.09%	49.85%	49.61%	49.37%	49.13%	48.89%	48.65%
	83%	33.44%	24.43%	50.69%	50.44%	50.20%	49.96%	49.71%	49.47%	49.22%	48.98%	48.73%	48.49%
	82%	33.03%	24.83%	50.56%	50.32%	50.07%	49.82%	49.57%	49.32%	49.07%	48.83%	48.58%	48.33%
	81%	32.63%	25.24%	50.44%	50.19%	49.93%	49.68%	49.43%	49.18%	48.92%	48.67%	48.42%	48.17%
													
	80%	32.23%	25.64%	50.31%	50.06%	49.80%	49.54%	49.29%	49.03%	48.78%	48.52%	48.26%	48.01%
	79%	31.82%	26.04%	50.19%	49.93%	49.67%	49.41%	49.15%	48.89%	48.63%	48.37%	48.11%	47.85%
	78%	31.42%	26.44%	50.06%	49.80%	49.54%	49.27%	49.01%	48.74%	48.48%	48.21%	47.95%	47.68%
	77%	31.02%	26.85%	49.94%	49.67%	49.40%	49.13%	48.87%	48.60%	48.33%	48.06%	47.79%	47.52%
	76%	30.62%	27.25%	49.81%	49.54%	49.27%	49.00%	48.72%	48.45%	48.18%	47.91%	47.63%	47.36%
													
	75%	30.21%	27.65%	49.69%	49.41%	49.14%	48.86%	48.58%	48.31%	48.03%	47.75%	47.48%	47.20%
	74%	29.81%	28.06%	49.56%	49.28%	49.00%	48.72%	48.44%	48.16%	47.88%	47.60%	47.32%	47.04%
	73%	29.41%	28.46%	49.44%	49.16%	48.87%	48.59%	48.30%	48.02%	47.73%	47.45%	47.16%	46.88%
	72%	29.00%	28.86%	49.32%	49.03%	48.74%	48.45%	48.16%	47.87%	47.58%	47.29%	47.01%	46.72%
	71%	28.60%	29.26%	49.19%	48.90%	48.60%	48.31%	48.02%	47.73%	47.43%	47.14%	46.85%	46.56%
													
													
													
													
													
	Gore	Revised	Did Not		Kerry Percentage of New Voters						
	Turnout	Weight	Vote	60%	59%	58%	57%	56%	55%	54%	53%	52%	51%
													
							Kerry Margin (millions)						
	100%	40.28%	17.58%	8.01	7.59	7.16	6.73	6.30	5.87	5.44	5.01	4.58	4.15
	99%	39.88%	17.98%	7.71	7.27	6.83	6.39	5.95	5.51	5.07	4.63	4.19	3.75
	98%	39.48%	18.39%	7.40	6.96	6.51	6.06	5.61	5.16	4.71	4.26	3.81	3.36
	97%	39.08%	18.79%	7.10	6.64	6.18	5.72	5.26	4.80	4.34	3.89	3.43	2.97
	96%	38.67%	19.19%	6.79	6.33	5.86	5.39	4.92	4.45	3.98	3.51	3.04	2.57
													
	95%	38.27%	19.60%	6.49	6.01	5.53	5.05	4.57	4.09	3.62	3.14	2.66	2.18
	94%	37.87%	20.00%	6.18	5.70	5.21	4.72	4.23	3.74	3.25	2.76	2.27	1.79
	93%	37.46%	20.40%	5.88	5.38	4.88	4.38	3.88	3.39	2.89	2.39	1.89	1.39
	92%	37.06%	20.80%	5.57	5.07	4.56	4.05	3.54	3.03	2.52	2.02	1.51	1.00
	91%	36.66%	21.21%	5.27	4.75	4.23	3.71	3.20	2.68	2.16	1.64	1.12	0.60
													
	90%	36.26%	21.61%	4.96	4.44	3.91	3.38	2.85	2.32	1.80	1.27	0.74	0.21
	89%	35.85%	22.01%	4.66	4.12	3.58	3.04	2.51	1.97	1.43	0.89	0.36	-0.18
	88%	35.45%	22.42%	4.35	3.81	3.26	2.71	2.16	1.61	1.07	0.52	-0.03	-0.58
	87%	35.05%	22.82%	4.05	3.49	2.93	2.38	1.82	1.26	0.70	0.15	-0.41	-0.97
	86%	34.64%	23.22%	3.74	3.18	2.61	2.04	1.47	0.91	0.34	-0.23	-0.80	-1.36
													
	85%	34.24%	23.62%	3.44	2.86	2.28	1.71	1.13	0.55	-0.03	-0.60	-1.18	-1.76
	84%	33.84%	24.03%	3.13	2.55	1.96	1.37	0.78	0.20	-0.39	-0.98	-1.56	-2.15
	83%	33.44%	24.43%	2.83	2.23	1.63	1.04	0.44	-0.16	-0.75	-1.35	-1.95	-2.54
	82%	33.03%	24.83%	2.52	1.92	1.31	0.70	0.10	-0.51	-1.12	-1.73	-2.33	-2.94
	81%	32.63%	25.24%	2.22	1.60	0.98	0.37	-0.25	-0.87	-1.48	-2.10	-2.72	-3.33
													
	80%	32.23%	25.64%	1.91	1.29	0.66	0.03	-0.59	-1.22	-1.85	-2.47	-3.10	-3.73
	79%	31.82%	26.04%	1.61	0.97	0.33	-0.30	-0.94	-1.57	-2.21	-2.85	-3.48	-4.12
	78%	31.42%	26.44%	1.30	0.66	0.01	-0.64	-1.28	-1.93	-2.58	-3.22	-3.87	-4.51
	77%	31.02%	26.85%	1.00	0.34	-0.32	-0.97	-1.63	-2.28	-2.94	-3.60	-4.25	-4.91
	76%	30.62%	27.25%	0.69	0.03	-0.64	-1.31	-1.97	-2.64	-3.30	-3.97	-4.64	-5.30
													
	75%	30.21%	27.65%	0.39	-0.29	-0.97	-1.64	-2.32	-2.99	-3.67	-4.34	-5.02	-5.69
	74%	29.81%	28.06%	0.08	-0.60	-1.29	-1.98	-2.66	-3.35	-4.03	-4.72	-5.40	-6.09
	73%	29.41%	28.46%	-0.22	-0.92	-1.61	-2.31	-3.01	-3.70	-4.40	-5.09	-5.79	-6.48
	72%	29.00%	28.86%	-0.53	-1.23	-1.94	-2.64	-3.35	-4.06	-4.76	-5.47	-6.17	-6.88
	71%	28.60%	29.26%	-0.83	-1.55	-2.26	-2.98	-3.69	-4.41	-5.12	-5.84	-6.55	-7.27
												
	

HERE'S A COMPREHENSIVE ELECTION 2004 SITE:
POLLING DATA, ANALYSIS, DISCUSSION
and...
THE EXCEL INTERACTIVE ELECTION MODEL
http://www.truthisall.net/

[PATRICK]

33. I think all the counter arguments

put the cart after the horse. The same forest of mess that obscures the complete verification of fraud, much less an accurate recount or retrieved votes from suppressed prevented voters, seems to validate faith in the result. That is the bias that seems to sneak into that attacks on the Kerry won crowd which slips too easily into the rationalizations for Dem "failure", always a mind and issues fog which means nothing to the hand that marks the ballot reality.

The vote cannot be verified. All agree that pre-stolen rigging and suppression was sufficient to switch the election. The vote will only be challenged locally and piecemeal but there is huge resistance to legally changing the the results and to the unwritten maxim it can't be done. The polls are fixed to the results, this time it would seem far enough beyond the "margin of error" to warrant a challenge- Ukraine style. GOP election fixing at its most blatant is barely prosecutable for the open crime it is and factors not at all into the media acceptance of the "contest" results.

TIA has every right and most of the convincing arguments while the opposition begins quickly to deny the right to challenge a secret election covered with signs of determined, no risk cheating at every single level- because it was necessary to beat the majority. There is nothing to do but support TIA even if one is not sure of the statistical proof, because not to do so validates that crime or an easier future one where navel gazing about issues and fighting the wrong battles is supposed to be the Dems lot in the new Empire.

Every claim is reasonable, obviously lacking destroyed evidence or any recourse. It is a demand for proof of validity for this criminal organization. It keeps us heading toward fair elections. The other way unnecessarily heads us back to defeatism or collaboration or silence because at root they fear the majority of Americans don't support Democrats(when it is PRECISELY those attitudes which continue to alienate a potentially crushing populist majority) and makes the dangerous parity where fraud can can be best hidden in the status quo. Perpetual sucker status and unverifiable elections leaves the people with wars and a trillion dollar debt so the over represented renegade rich can get subsidies from our forced support of a civil government which is not allowed to function. To buy into a Bush victory is to drink deep of the deception MOST Americans heartily resist without being able to so well grasp the situation as activists here. I think a higher percentage of Dem centrist politicians buy into the Beltway media line more than any of their real constituents. They live in a statistical Never Never Land before one vote is actually counted- or disappeared.

[tn-guy]

34. Here is where the wheels fall off the argument

I'm going to try to lay out my "counter argument" and the assumptions I use. I would welcome someone showing me the fallacy in my reasoning. Calling me a freeper, or a naysayer or some other name is not a valid counter argument. However; if I misstate TIAs arguments or have a logical error please let me know.

1. Statement - The TIA argument is based on using the results of the NEP to show that the election was stolen.
2. Assumption - For that argument to make sense, one must believe in the validity of the NEP. Otherwise one is basing an argument on bogus data.
3. Statement - As part of TIA's argument to demonstrate a stolen election, TIA proves that the results for the "who did you vote for in 2000?" question cannot possibly be correct.
4. Assumption - Proving that part of the NEP results cannot possibly be correct tends to undermine the validity of other parts.
5. Assumption - When conducting serious research one should not cherry pick those parts of a flawed survey that tend to support ones' position while discarding those parts which do not - especially when one has already demonstrated that the survey is questionable at best.
6. Assumption - Since TIA has demonstrated conclusively that part of the NEP survey results are flawed, then the entire survey is suspect.
7. Conclusion - Therefore; any conclusions that rely on the NEP are similarly suspect. In other words, garbage in - garbage out.

I'm not trying to be difficult here. It's just that I think it would be a more productive use of everyone's time to focus on the future rather than a futile attempt to "prove" something based on bogus data.

[LeftCoast]

37. Excellent post

Looking forward to the response.

[autorank]

38. Channeling TIA:

Post 32 needs to be read carefully, which is important to understanding the arguments below.

1. Statement - The TIA argument is based on using the results of the NEP to show that the election was stolen.

TIA: You fail to grasp the mathematical logic.

The FINAL NEP HAD to use MATHEMATICALLY IMPOSSIBLE weightings for it to MATCH TO THE VOTE and so Bush won the Final NEP, 51-48.

Kerry won the earlier timelines, in which the weights were closer to reality. Look at the numbers. Do the calculations. Can you derive the national vote shares based on the demographic category weightings and percentages? You will understand the logic only if you confirm the calculations displayed in post # 32

2. Assumption - For that argument to make sense, one must believe in the validity of the NEP. Otherwise one is basing an argument on bogus data.

TIA: On the contrary, the fact that the FINAL NEP is bogus means that Kerry won the election. Just plug in the appropriate weightings. It’s all there in post # 32.

3. Statement - As part of TIA's argument to demonstrate a stolen election, TIA proves that the results for the "who did you vote for in 2000?" question cannot possibly be correct.

TIA: That is only partially correct; you need to be more precise. You agree that FINAL 13660 at 1:25pm is impossible and cannot possibly be correct. Do you also agree that since it matched the vote count (51-48(, then the vote count also must have been incorrect? The earlier NEP timelines (4pm, 7:30pm and 12:22am) all won by Kerry 51-48, were therefore close to the truth. Once again, I draw your attention to post # 32.

4. Assumption - Proving that part of the NEP results cannot possibly be correct tends to undermine the validity of other parts.

TIA: That is not true. It ENHANCES them. With plausible weights, Kerry wins all scenarios. To repeat, the FINAL NEP weights had to be adjusted to 43/37 (mathematically impossible) in order to match the recorded Bush vote, 51-48. You can see that Kerry wins all PLAUSIBLE scenarios in post # 32.

5. Assumption - When conducting serious research one should not cherry pick those parts of a flawed survey that tend to support ones' position while discarding those parts which do not - especially when one has already demonstrated that the survey is questionable at best.

TIA: Cherry-picking what? You are grasping at straws here. Name one “part” which has been discarded, other than the IMPOSSIBLE 43/37 weights. Judge for yourself. Take a look at post # 32.

6. Assumption - Since TIA has demonstrated conclusively that part of the NEP survey results are flawed, then the entire survey is suspect.

TIA: No, only the FINAL Exit Poll at1:25pm which matched to the Bush vote, is suspect. See why in post # 32.

7. Conclusion - Therefore; any conclusions that rely on the NEP are similarly suspect. In other words, garbage in - garbage out.

TIA: The “garbage” was the 43/37 weights. So I took them out, and replaced them with mathematically feasible weights. Check out the sensitivity analysis of Bush/Gore voter turnout and New voters on the popular vote in post # 32

You: I'm not trying to be difficult here. It's just that I think it would be a more productive use of everyone's time to focus on the future rather than a futile attempt to "prove" something based on bogus data.

TIA: Those who ignore history are doomed to repeat it. Carefully study the evidence in post #32.

[tn-guy]

39. My bad......

Apparently, I'm missing something obvious here and I'm also growing weary of this so this will probably be my final posting on this thread. I'm not going to address this point by point but I will try to show how confused I am.

From post #38

Post 32 needs to be read carefully, which is important to understanding the arguments below.

1. Statement - The TIA argument is based on using the results of the NEP to show that the election was stolen.

TIA: You fail to grasp the mathematical logic.

The FINAL NEP HAD to use MATHEMATICALLY IMPOSSIBLE weightings for it to MATCH TO THE VOTE and so Bush won the Final NEP, 51-48.

Kerry won the earlier timelines, in which the weights were closer to reality. Look at the numbers. Do the calculations. Can you derive the national vote shares based on the demographic category weightings and percentages? You will understand the logic only if you confirm the calculations displayed in post # 32

From post #32


"We will show that National Exit Poll Bush 2000 weights are
mathematically
IMPOSSIBLE in both the preliminary and final timelines."



I guess my misunderstand comes from thinking that when one says the weights are mathematically impossible in both the preliminary and final timelines one means both "preliminary" and "final". It seems that when one says "both preliminary and final" one actually means only "final" were mathematically impossible and "preliminary" weights were "closer to reality." A further source of confusion is that it seems to me if "preliminary and final" results are mathematically impossible then one should not use either "preliminary and final" results for any further analysis. To put it another way, once we have determined that the NEP "cooked the books" I am unwilling to trust ANY numbers in the NEP. I think deciding to trust some numbers and not trust others is the definition of cherry-picking data.

Of course, I'm not a professional statistician so I'm hardly an expert. However; I am by education and profession and engineer so I am used to numbers and math and sampling and quality control and stuff like that. Were someone to give me a data collection as internally inconsistent as the NEP results when I was trying to diagnose a production problem I'd toss the whole set out. When a set of data points are internally inconsistent one only knows that they are questionable, to start assuming which data are good and which are bad will put you on dangerous ground quickly.

[autorank]

40. Whatever...

" I guess my misunderstand comes from thinking that when one says the weights are mathematically impossible in both the preliminary and final timelines one means both "preliminary" and "final"

This research is necessarily ad hoc. We couldn't construct a set of scenarios that would make it as pristine as it might be, given that it's obviously not a laboratory scenario. Nevertheless, there is enough there to work with and there is a robust debate on these issues. The inferences made from the data regarding inconsistencies revolves mostly on the next day poll when the results were retrofitted to meet the election results, not a poll, as the first four were individually and aggregated, but more of a political statement, I suspect. Thus the concentration on the first four. These match well with the state exit polls which have a n of around 70,000 (working from memory) and have a killer MoE. This can be debated endlessly but the combination of the readily available data plus the general atmosphere of a stolen election in 2000, no doubt about that, and the general pattern of what's fair to call criminal behavior by the Bush administration give great weight to the interpretation that points to fraud. Besides, given the history of American elections, it's best to question first rather than assume a "clean" election. Read Deliver the Vote if you have any questions about our history.http://www.amazon.com/gp/product/078671591X/104-6392133-1527111?v=glance&n=283155

[PATRICK]

41. That seems to point out

exactly what I was saying in support of TIA, as little as I understand of the number crunching. The most powerful "weighting" in TIA's favor is the consistency, the necessity, the predictability and the MAMMOTH research into methodologies of fraud employed. It must be frustrating trying to maintain some belief that Bush won when all the proof must be destroyed or changed or mystified in the process- such as granting that the NEP was "erroneous" and either must be corrected or dismissed.

Tell that to Putin and the Ukraine. They WISH they had that kind of control over the election. In essence, once resorting to trashing all the NEP in order to remove TIA from the board, one is left with the feeble fabrications of "morals" analysts who ALSO have cooked things from PARTS of the cooked NEP to have it their way. No one even talks about that embarrassment anymore. I wonder why? Instead a more shadowy presumption moves underneath the Kerry kicking, shoot-our-loser talk. Rove, that guy Rove, who told us how he did this clever thing here, that clever thing there. Wow, I guess we just have to do things better, maybe like not putting so many big words in the platform and toning down the populism. Oh yeah, Bush was such a great war president and blah blah blah the terrified sheeple always veer toward the lucky faux warlord. Anyone buying that one now? I for one would like to at least keep the process of elimination going over our shadow president's second "victory" over the American voter. Or maybe there are millions of secret Amish voters hidden in farm compounds scattered around the key state precincts.

And never forget, even though the votes cannot be retrieved, the fraud, the perps, the witnesses, and evidence of fraud CAN. In fact, there is an inexorable march toward nearly all these things and people who would nervously dismiss the very idea of prosecuting this buried issue. None of Bush's advantages that one single debate performance in a fair media would have sunk like a stone, do more than explain why Kerry didn't sweep the board, millions of votes suppressed or not.

[OnTheOtherHand]

43. well, I hope you stop by to pick up your fan mail

TIA's argument isn't as bad as you might think -- his objection, that the raw data cannot be reconciled with the official returns unless one stipulates many more Bush 2000 than Gore 2000 voters, has some force. Incidentally, political scientists quickly get used to taking garbage data, spitting on it and wiping, then proceeding to analyze it. (Seriously, we don't have much choice. We can't, for instance, reengineer our respondents to give straight and accurate answers; we just do what we can to minimize, to compensate for, and to be wary of bias.)

Unfortunately, what TIA "misses" is survey after survey after survey that documents false reporting of past vote. For instance, I recently examined the data from the Pew Research Center's last pre-election survey. Among registered voters who recalled a vote choice, Bush beat Gore by 14 points. That's extreme, but you can head over to the General Social Survey -- umm, actually the site seems to be down right now, but usually this works: http://www.icpsr.umich.edu:8080/GSS/homepage.htm -- and run crosstabs that show incumbents running up the score in later years. I think Dukakis loses by at least 30 points before they stop asking about him. Poor guy. Yes, he looked silly in the tank, but his campaign wasn't that bad. (Nixon is the exception, sort of: he only more or less breaks even against McGovern.) Same thing in the National Election Studies. I can do it with past exit polls, too.

TIA has been told this, but he doesn't deal well with discrepant information IMHO. And he can't even find a stake to burn me at. So we're pretty much stuck at the epicycle stage.

[Yupster]

100. By the way Patrick

the final exit poll in 1988 had Dukakis and Bush tied even though Bush actually won by a pretty solid 8 %.

Which just proves that the exit polls are never wrong.

[Al-CIAda]

42. .

[autorank]

44. The Good News! False recall/polls cleared up; not relevant.

The issue of Gore voter “false recall” is a STRAW MAN argument meant to divert us from the FACTS. It’s not even the issue here. IT IS IRRELEVANT TO THE MATHEMATICAL ARGUMENT PRESENTED. Let’s put it to rest once and for all.

Gore/Bush voter weights in the 2004 NEP are CONSTRAINED by the number of ACTUAL 2000 voters still alive to vote in 2004. Why would anyone insert the issue of "FALSE RECALL" when we're talking about actual votes?

Of those who voted and whose votes were reported in 2000, how many survived until 2004 and how does that impact on the assumptions in the National Exit Poll?

FALSE RECALL IS NOT AN ISSUE. ELECTION RESULTS EXTRAPOLATED AND CARRIED FORWARD IS THE FOCUS OF THIS ANALYSIS.

THE NUMBER OF GORE VOTERS AND BUSH VOTERS STILL ALIVE TO VOTE IN 2004 IS THE ONLY ISSUE.
autorank

Channeling TIA:
TIA:
Once again, you fail to acknowledge the basic facts.
So here we go again.

What were the MAXIMUM number of Bush 2000 voters who COULD HAVE VOTED IN 2004?
That number is 48.7 million. PERIOD.
Do you accept that as a FACT?

There is NO other stipulation.

So the NEXT QUESTION for you is simply this:
WHERE DID BUSH GET 13 MILLION VOTES TO GET TO 62 MM FROM 48.7 MM?

You better look at post # 32 again.

HERE IS THE "CLINCHER" CHALLENGE :
USING THE FACTUAL NEP DATA IN POST #32, CAN YOU COME UP WITH
ONE SOLITARY, FEASIBLE, PLAUSIBLE, REALISTIC BUSH WIN SCENARIO?

IF YOU CANNOT PROVIDE ONE, THE GAME IS OVER.

Excellent historical volume that will explain why we should begin with questions on any election that raises suspicions. Time to give up the notion that we start with the presumption of innocence when it comes to suspicions of election fraud

[OnTheOtherHand]

45. oh brother, here we go again

Obviously the number of surviving Bush and Gore voters is not the only issue. Although if we can stipulate that, it will save me the trouble of actually reading #32. I would like that very much.

Bush got the remainder of his votes from people who didn't vote for him in 2000 -- same as in the pre-election polls.

If TIA thinks that false recall (or false reporting or whatever one wants to call it) is a red herring, it must be because he never read or understood my posts the last time. It affects all the other results.

Here is an approximation that is consistent with the facts known to me:

The 2004 presidential electorate consisted of some 122.3 million voters, of whom

46.3 million had voted for Gore, and  6.8 million of these voted for Bush
45.8 million had voted for Bush, and 42.3 million voted for Bush again
 3.7 million had voted 'other',  and  0.7 million of these voted for Bush
26.5 million did not vote,       and 12.3 million of these voted for Bush

That's Bush's official vote within rounding error. Note that I came in well under the 48.7 million Bush voters of which the preceding post states, "There is NO other stipulation." Heh.

Obviously I cannot tell how close these numbers are to being accurate. But if someone comes back and tells me, "But we know that a much smaller percentage of Gore voters voted for Bush in 2004," I will point out again: that assumes not only that the exit poll sample is approximately fair, but also that people who say they voted for Gore in 2000 actually did, and people who say they voted for Bush actually did. That's why false recall matters.

[autorank]

46. Oh Brother, where art thou?

A definitive scenario for a Bush victory has been
provided.Since you left out the demographic percentages, I've
calculated them so that we can investigate your analysis in
relation to NEP polling data.

I’m happy that you agreed to the mathematical stipulation that
the maximum Bush 2000 voter turnout was 48.7mm. In so doing,
you have accepted that the Final Exit Poll weightings of
43/37% Bush/Gore are impossible. So we have a legitimate basis
for further analysis. 

The question is: where do you go from there? How do you derive
the Bush 3 million vote margin, while adhering to the
weighting constraints? Well, the first step is to assume
feasible turnout weightings for Bush and Gore voters. You have
done so. Your weightings are reasonable, even somewhat
conservative. 

Since the playing field has been leveled as far as the
weightings are concerned, we can now look at actual vote share
scenarios. This is where we part company.

To believe your Bush 3mm win scenario, we must believe the
following:
1- One out of seven (14.6%) Gore 2000 voters switched to Bush.
2- Only 6.6 percent of Bush 2000 voters switched to Kerry.
3- Kerry won 52.6% of those who did not vote (DNV) in 2000, 
These are New voters and others who voted prior to 2000, but
sat that one
out.

The vote share assumptions are arbitrary and have no
evidentiary basis. They are far removed from the National Exit
Poll at 12:22am (13047 respondents) and the 1:25pm final
(13660). 

These are the corresponding percentages:

..........................12:22am 1:25pm OTOH
1. Gore voters for Bush:      8;  9;   14.7
2. Bush 2k voters for Kerry: 10;  9;    6.6
3. DNV for Kerry:            57; 54;   52.6 

Considering that the exit poll MoE is 1% for the national vote
and 2% for sub-samples, the probabilities of your vote shares
are remote. You had to assume that Bush did significantly
BETTER than he did in the final exit poll, which he won 51-48%
and was matched to the recorded vote. 

You based your original case on the “false recall”
hypothetical: that Gore voters, but not Bush voters, forgot
who they voted for in 2000. This was your rationale in order
to explain the 43 Bush/37% Gore weightings. As I stated
earlier, the "false recall" argument is irrelevant
and moot - now that you have agreed that the 43/37% weightings
were impossible to begin with. You use feasible, plausible
weightings in your Bush win scenario. 

Time for a reality check. 

How is one to believe your assumption that 1 out of 7 Gore
voters who saw the election stolen from them by Bush in 2000,
would forgive or forget and vote for Bush in 2004, especially
when his job performance has been, shall we say, less than
satisfactory? 

He had a 48.5% approval rating on Election Day. 

Furthermore, Democratic 2004 voter registration far exceeded
the Republicans in spite of massive evidence of voter
disenfranchisement, especially in the key states of Florida
and Ohio.

Finally, are we to believe your claim that Bush won by a 3
million vote margin? Because that is EXACTLY WHAT YOUR
IMPLAUSIBLE SCENARIO SUGGESTS. 

In other words, are you in fact telling us FRAUD WAS NOT
COMMITTED IN THE ELECTION? Because that is what your scenario
shows.  Difficult to imagine.

The challenge has not been met.
The Bush win scenario is IMPLAUSIBLE. 
It doesn't pass the smell test.

Here are calculations of national vote shares, assuming your
weightings
for: 
1) your Bush win scenario 
2) the 12:22am exit poll 
3) the final 1:25pm exit poll

Assuming your (reasonable) weights, Kerry handily wins both
Exit Poll
timelines. 

OTOH 				 Bush winning margin: 3.123mm
	Votes	BushV	BushP	Kerry	Kerry%	Other	Other%
DNV	26.5	12.30	46.4%	13.9	52.6%	0.3	1%
Gore	46.3	6.80	14.7%	39.0	84.3%	0.5	1%
Bush	45.8	42.30	92.4%	3.0	6.6%	0.5	1%
Other	3.7	0.70	18.9%	3.0	80.1%	0.0	1%
							
Total	122.3	62.10	50.78%	59.0	48.22%	1.2	1.00%

12:22am NEP (13047)		Kerry winning margin: 7.879mm
	Votes	BushV	BushP	Kerry	Kerry%	Other	Other%
DNV	26.5	10.87	41%	15.11	57%	0.27	1%
Gore	46.3	3.70	8%	42.13	91%	0.00	0%
Bush	45.8	41.22	90%	4.58	10%	0.00	0%
Other	3.7	0.78	21%	2.63	71%	0.30	8%

Total	122.3	56.57	46.25%	64.45	52.69%	0.56	0.46%


1:25pm Final NEP (13660)	Kerry winning margin: 3.719mm
	Votes	BushV	BushP	Kerry	Kerry%	Other	Other%
DNV	26.5	11.93	45%	14.31	54%	0.27	1%
Gore	46.3	4.63	10%	41.67	90%	0.00	0%
Bush	45.8	41.68	91%	4.12	9%	0.00	0%
Other	3.7	0.78	21%	2.63	71%	0.30	8%
							
Total	122.3	59.01	48.25%	62.73	51.29%	0.56	0.46%

[autorank]

47. FIN

[OnTheOtherHand]

48. I hope "FIN" means that TIA is about to give this up

although I am not very optimistic.

I think it's meaningless to say that "the Final Exit Poll weightings of 43/37% Bush/Gore are impossible" (I mean, what? the computers blow up or something?), but I'm happy to stipulate that 43% of 2004 voters didn't vote for Bush in 2000.

As I explained at length months ago, my vote share assumptions are grounded in actual data from a 2000-2004 panel study. They reflect actual people's actual reports in 2000 of whom they had just voted for, as well as their actual reports in 2004 of whom they had voted for in 2000 as well as in 2004. Of course I don't think that the people in this panel are representative of the people in the exit poll. Given panel attrition, I would expect the people who made it through the panel to be more politically aware overall than the exit poll participants, and therefore somewhat less likely either to misreport their past votes or to defect from their 2000 choice. But I could be wrong. (EDIT: Of course another thing about the panel is that the respondents presumably overreported having voted at all. I haven't sorted out yet how that might affect my estimates -- e.g., whether maybe I should postulate less Gore'00->Bush'04 and more DidNotVote'00->Bush'04 -- but it certainly doesn't make the exit poll vote'00 question any more useful as proof of fraud.) Regardless, very little in my scenario is just made up.

TIA claims both that false recall doesn't matter, and that it contradicts the final weightings for me to say that 14+% of Gore voters defected. This seems to be invincible confusion on his part. If 14.6% (or whatever) of actual Gore 2000 voters voted for Bush, but many of them in 2004 report being Bush 2000 voters, then obviously (isn't it obvious? why isn't it obvious?) they won't show up in the exit poll as having been Gore 2000 voters, whether or not the exit poll is weighted to 2004 results.

Since there is no likelihood that TIA will understand this, say, the twentieth time I explain it as opposed to the nineteenth, I encourage anyone who has questions about it to PM me. Otherwise, I will assume that it is just the sort of thing that either one gets, one doesn't, or one doesn't care about either way.

Historically, a 48.5% approval rating appears good enough to win by more or less the margin that Bush won by. (This particular plot uses last available Gallup approval ratings; I've tried alternative specifications, haven't seen much difference.)


Since TIA presents no evidence about registration figures, there's nothing to refute there.

Then comes what TIA apparently regards as his definitive smackdown:

Finally, are we to believe your claim that Bush won by a 3
million vote margin? Because that is EXACTLY WHAT YOUR
IMPLAUSIBLE SCENARIO SUGGESTS. 

In other words, are you in fact telling us FRAUD WAS NOT
COMMITTED IN THE ELECTION? Because that is what your scenario
shows.  Difficult to imagine.

The challenge has not been met.
The Bush win scenario is IMPLAUSIBLE. 
It doesn't pass the smell test.

Short TIA: "I JUST DON'T BELIEVE that Bush won by 3 million votes." Umm, OK. That would have been a much shorter OP. Even shorter would have been, "I'm not LISTening! You can't MAKE me!"

[Febble]

49. My two pence

I notice that TIA is happier with the column graph function in Excel that with the scatterplot function, although I do admire the splendid time-series scatter-plot he does with Bush's approval rating, 4th degree polynomial fit and all, so I know he knows how it works.

However, I do suggest he plots a few LINEAR regressions through some of his data, which may help him understand some of the points OTOH has been making.

One of which is that the "incumbent rule" appears to mean that there is a strong linear relationship between an incumbent's approval rating and his margin in his re-election vote, as TIA correctly implies. However, OTOH's plot indicates that the break-even point is not 50% as often alleged before the election, but something nearer to 43% (eyeballing OTOH's plot with the side of a piece of paper and a ballpoint). To be fair, we should replot this without 2004 in order to get a predictive value for 2004, but as the 2004 datapoint is fairly close the regession line it won't make a lot of difference.

It would also be possible to work out the MoE for that break-even point, something that TIA might like to do, as MoEs are his MO it seems (nerdy pun). But the MoE will certainly include Bush's margin in 2004, something I think all we quanties can see by looking, presumably including TIA, but a number would be nice.

And when he's done that, TIA might like to plot the correlations between the DIFFERENCES between the pre-election polls and the vote-count on one axis, and the DIFFERENCES between the exit polls and the vote-count on another axis, and STARE at the results, preferably with the aid of a LINEAR best fit line plotted through them (NO polynomials IF you please, thankyou), and decide what it means for the BEAST in the EAST.

And as a 3rd pence, I will say that the fact that neither I nor OTOH are convinced that the exit polls are evidence for fraud does NOT mean that either of us think that no fraud occurred, or that the election was fair, or that the voting machines can't be hacked, or that having machines in which no-one can be sure their vote was counted is a good thing. Personally, I think voting on DREs is mad (=crazy in English English). Also I think the level of voter disenfranchisement, especially of ethnic minority voters, is a scandalous Civil Rights violation. So before anyone starts on the "Naysayer!" schtick, I'll say here and now: for the so-called leader of the free world, American democracy sucks. It needs radical fixing, which means an independent media, restrictions on campaign funding, universal franchise, transparent auditable voting systems and a secret ballot. Probably some other stuff too.

To have the debate dominated by whether or not a survey was more than a few percentage points off or not, when there is copious evidence to show they can be and that this one probably was, seems to me to be nuts. But as long as people keep insisting that the exit polls show that the 2004 election was stolen, I'm going to keep saying they don't. Because they don't. It could have been. But the exit polls don't show that it was. If anything, they tend to show that if it was, it wasn't done by vote-switching. Personally I think it was done by the simple expedient of making sure that those with most to gain from a Kerry victory, and most to lose from a Bush one, didn't get to vote.

[OnTheOtherHand]

50. as usual, your post rewards close reading

and I am sure that your dissertation will as well. ;)

There does seem to be a tedious rhetorical waltz that goes ONE-the-exit-polls-prove-fraud TWO-actually-they-don't THREE-why-are-you-defending-war-criminals? Hard to dance to that. Probably doesn't do much for the cause of election reform, either.

Now I'm afraid that someone is off saying that "In order for the official totals to be accurate, Bush had to have won 14.6% of the Gore 2000 vote." No, there are lots of smaller numbers that could work, too. It would be possible to enter into a serious discussion of what numbers make the most sense (not ruling out the possibility of vote miscount). But it's probably hard to do that if one is convinced from the get-go that Kerry won by several million votes, and is just trying to "prove" it.

Your post points to several other interesting conversations -- some of them quite important -- that probably won't happen here either. Oh well.

[anaxarchos]

51. Forgive me for interrupting….

…your mutual admiration society.

I want to thank you for making a strong argument for the position that GW Bush could not have won the popular vote in the 2004 presidential elections without significant fraud. I understand that you would like us to come to the opposite conclusion, but by proving the difficulty of coming to that conclusion, you have done us a rare service. Your argument comes in two parts, both of which underline the arithmetic of 2004:

Of the few potential scenarios for a “legitimate” Bush victory, you have chosen the one which, while difficult to reconcile, is more "plausible" than the others. The fabric of the 2004 presidential elections was not particularly elastic. The alternatives were a swing of new voters to Bush instead of to Kerry, Republicans turning out in record numbers while Democrats "sat out the election", or the wholesale rush of Nader voters to the Republican Party. Obviously, these alternatives contradict a good deal of existing evidence to the contrary.

Instead, you have chosen the "Gore voters defected to Bush at two times the rate that Bush 2000 voters defected to Kerry". In support of this premise, you have offered no evidence whatsoever. Instead, you have challenged evidence to the contrary by arguing that the rate of defection indicated in the exit polls (fairly similar for both candidates) is suspect because of "false recall". In the main, you have made an argument similar to rBr, i.e. that seemingly impossible splits in the exit polls can be explained by response error.

You have also thrown a few "stinky fish" on the table which I take to be red herring. I believe your usage of your panel studies, your tangential discussion of incumbent approval rates and your generic brand of “false memory” are all downright silly. But, I also have experience with how quickly such objections take the debate down a rat hole. I will let them stand for the time being. Instead of challenging these, I will try to stay “on point”.

So too with your “point shaving”. I won’t quibble with your numbers at the moment despite the fact that you have made no argument as to why the Bush defection rate should be reduced (without this, the Gore 2000 defection rate approaches 18% according to you). On the contrary, should your premise prove true, the corollary is that the Bush defection rate is understated by the exit polls.

The same goes for evidence outside of the context of the exit polls themselves. It doesn’t seem to trouble you that only you have detected this mass defection. It seems to have been missed by pre-election pollsters. It seems to have been dismissed by both major parties. It also seems to have been ignored by post-election researchers despite the availability of more direct evidence than polling data. If I am mistaken on this, please provide references. In the meantime, I will defer my objections until you have actually made a case.

Instead of these possible distractions, let's explore "Gore voters defected to Bush at twice the rate that Bush 2000 voters defected to Kerry" in the context of the exit polls alone.

Of all of the survey questions, you want to eliminate the most directly pertinent one on the basis of possible response error. Even if we accept that, the retrospective question is by no means the only relevant question. There are many others.

There is, for example, the party ID question. I understand that this question is controversial because it may be one of the splits that were altered when the 2004 exit polls were “reconciled” with the actual vote count. But that controversy should not impact the internal “defection rate” reported by the exit polls. According to the 2000 exit polls, 11% of respondents who identified themselves as “Democrats” voted for GW Bush. According to the 2004 exit polls, the defection rate was an identical 11%. There is no indication of increased defections here.

To test this in another way, we can look at the internal defection rate of respondents by self described “ideology”. In 2000, 80% of those who describe themselves as “liberal” and 52% of those who describe themselves as “moderate” report that they voted for Gore. Together, they make up 90% of the Gore vote. In 2004, 85% of those who describe themselves as “liberal” and 54% of those who describe themselves as “moderate” report that they voted for Kerry. Again, there is no indication of increased defections.

The most decisive of the survey questions, however, are the two GOTV questions. These are the only ones in which respondents are not asked to report on themselves but to report on how others characterize them.

For the survey question, “Were you contacted by the Kerry campaign?”, 26% answered yes, of whom 66% reported voting for Kerry. To the matching question, “Were you contacted by the Bush campaign?”, 24% answered yes, of whom 62% reported voting for Bush.

Let me assure you that neither political party suffers from “false memory”, at least as far as GOTV efforts are concerned, and that both are quite proficient at identifying their base. That is precisely who was “contacted”. In addition, there is likely to be relatively little overlap between these populations which will consist primarily of Gore and Bush voters from 2000. There is, once again, not only no support for your defections premise but, on the contrary, an indication that the Democratic GOTV did a slightly better job and achieved a slightly better hit rate.

Locally, these same GOTV questions shed even more light. In the Florida 2004 exit polls, for example, 35% report being contacted by the Kerry campaign while 34% report being contacted by the Bush campaign. This increased GOTV effort is clearly a byproduct of Florida’s battleground status, but, because of its vastly increased voting population in 2004, the two Florida GOTV populations in the aggregate represent virtually all of the voters from 2000. Yet again, there is no indication of varying defection rates between the two parties. The Kerry hit rate was 65% while the Bush hit rate was 63%.

Where did you come up with this theory of yours? It doesn’t “work” at all, let alone with “much smaller numbers”. I’m beginning to wonder if you, like our friend from Drexel, are merely giving a “scientific” voice to press accounts. But few of those accounts deserve any standing. Consider the story that “Suburban Soccer Moms”, presumably Gore voters in 2000, became “Security Moms” in 2004 because of the last minute release of the Bin Laden tape. This story was even given currency by Kerry himself. But, were those who made up Bush’s “margin of victory” really transformed “Suburban Soccer Moms”?

Suburban?

According to the 2000 Exit Poll survey question, “Where do you live?”, 43% of respondents reported that they lived in the suburbs and 47% of these voted for Gore. In 2004, 45% reported that they lived in the suburbs and 47% of these voted for Kerry. Nope, they weren’t “suburban”.

Moms?

According to the 2000 Exit Poll Survey Question, “Do you have children under 18?”, 39% answered “yes” and 45% of these voted for Gore. In 2004, 37% answered “yes” and 45% of them voted for Kerry. And, of course, that includes both “moms” and “dads”. Nope, they weren’t “moms”.

Worried about the Bin Laden Tape?

Asked to characterize the importance of the Bin Laden videotape, 32% of respondents in the 2004 exit polls said they were “very important” but, of these 53% reported that they voted for Kerry. Considering that nearly 90% of all respondents also reported having made their decision before the tape was released, there is no case here. Nope, it wasn’t the tape.

The press accounts of “suburban soccer moms” turning into “security moms” in order to re-elect GW Bush were attempts to explain by anecdote what is very difficult to reconcile on the facts. In the end, they were fables… much like your story, I’m afraid.

[OnTheOtherHand]

52. I stopped on the first egregious factual error

Instead, you have chosen the "Gore voters defected to Bush at two times the rate that Bush 2000 voters defected to Kerry". In support of this premise, you have offered no evidence whatsoever.

My evidence is that in the 2000-04 NES panel, Gore voters (as self-reported in 2000) did defect to Bush at two times the rate that Bush 2000 voters defected to Kerry. I realize that the NES panel isn't directly comparable to an exit poll, but if memory serves, you haven't even tried to offer an alternative account.

I have no idea why you are apparently as clueless as TIA about (1) the evidence that polls quite often exaggerate past voting for the incumbent and (2) why this would affect the polls' (including but not limited to exit polls') estimates of defection rates.

You seem to want to be taken seriously -- and if I find something in here that can be taken seriously, I will get back to you.

[anaxarchos]

65. The NES provides you no help here at all...

... for reasons we have already discussed many times. I have also cited in the past, several NES technical reports which directly contradict your attempted usage ("no you haven't!"... "YES, I have!", etc..).

1) Since you stopped reading, you might not be aware that my post focused (with considerable self-discipline) on evidence intrinsic to the exit polls. I also do not believe that you have presented any evidence extrinsic to the polls (unless we want to debate whether anecdotal evidence is "evidence") but that is not at issue here. One thing at a time.

2) This is a "factual error"? "Egregious", yet? Shit man, you don't belong in politics...

3) I repeat, the exit polls DO NOT support your theory of an increased Kerry defection rate hidden by response error on the retrospective question.

4) Is it your position that the party ID, political ideology, AND the GOTV questions are also subject to an incumbency effect? If so, not only do you contradict the literature but you have just declared all political polling immaterial.

5) In the past you have claimed that TIA engaged in childish denial. Take a look in the mirror: "I stopped on the first egregious factual error", "you are apparently as clueless as TIA", "You seem to want to be taken seriously", "if I find something in here that can be taken seriously, I will get back to you".... How is this different from "Nyah, Nyah... I'm not listening. I'm not listning... Look, I have my fingers in my ear"?

6) Lest I forget, I claim the point by right of "truth and virtue".

Luv-40

[OnTheOtherHand]

87. whoops, missed this one

No, I don't agree at all. The NES panel shows people who say in 2000 that they voted for Gore, deciding by 2004 that they had voted for Bush (and fewer going the other way). And the panel shows these people skewing the defection rates. You have never explained how vote overreporting creates this effect, or whatever you think it does; you simply use it as an excuse to disregard the discrepant information.

I have no idea why you believe that the party ID, ideology, and GOTV questions have any bearing on the Gore defection rate, so I don't come to the question of "incumbency effects." But if I did, and if I concluded after analysis that there was such a thing, it would still not declare all political polling immaterial. You know better than that. You know you do.

#84 is not a snit; it is my current considered opinion.

[Febble]

53. And here is Thing Two

come to join in.

You are arguing (as I argued myself many moons ago) that Bush couldn't have won without fraud, because he couldn't have attracted the support required to win. That's a fair enough argument.

So you turn to the exit poll evidence, which, if you believe the above, appears to provide supporting evidence that there was fraud, because the exit poll discrepancy cannot have been due to chance, and because the reweighted cross-tabs on recalled vote cannot possibly indicate respondents' actual vote in 2000.

But there is an alternative interpretation of the exit poll data, which is, as both OTOH and I have said, that the discrepancy between the poll and count was due to polling bias, and that the reweighted 2000 vote proportions reflect people's tendency, as evidenced by the the panel study he cites, to report having voted for Bush when in fact they voted for Gore.

You are, of course, welcome to choose to stick with the first interpretation. We, presumably, are free to stick to the second, although I have no intention of sticking with any interpretation if it is undermined by other evidence. Because the only objective way to disambiguate a finding is to look at other evidence.

And (like wli on PI), the evidence I've been looking at is correlations between redshift and various other stuff, and so far it's not convincing me that the exit poll discrepancy had much to do with vote-switching fraud. Because, if vote-switching fraud is the latent variable that underlies redshift in the exit poll, redshift in the pre-election poll, and redshift relative to 2000 vote (aka "swing") - in other words if the vote-count was redder than it should have been because of fraud - all these redshift measures should be correlated with each other. And they aren't.

Therefore, it would seem, vote-switching fraud is unlikely to be the latent variable that underlies those three sets of redshifts. This raises the possibility that the cause of the poll discrepancies is something else, e.g. polling bias, which you would expect to be independent of each other, as they are conducted by different organisations. At which point, the Gore proportions in the final exit poll cross-tabs are only a bar to the polling bias explanation if you refuse to accept the panel study evidence that false recall, in the direction required to explain the reweighted cross tabs, occurs (or, as some would have it, that polls are ever biased :rofl:)

To me, the most subtantial body of evidence regarding electoral injustice in 2004, is the evidence that as usual, a substantial swathe of voters, disproportionately African American, and disproportionately Democrat, were yet again systematically disenfranchised in all sorts ways.

And because this stuff has been going on for years, it would not surprise me if it contributed to the exit poll discrepancy and to Bush's win, yet didn't show up in a swing-shift correlation in the exit polls. It would also not surprise me if it was actually greater in precincts with the kinds of voting technology (punchcards, levers) that in the past have been associated with high residual vote rates, not with DREs (although one study I did with Josh Mitteldorf suggested that push-button DREs were associated with greater rates of undervotes in AfricanAmerican and Hispanic precincts than touch-screen DREs, and of course I firmly believe that DREs resulted in a large net loss of votes to Kerry in Ohio, simply because they were inequitably distributed).

So why don't we call a truce in this really very silly war over exit polls - which may yet tell us where the bodies are buried, but which won't if there is an a priori refusal to consider any interpretation of the data that conflicts with the premise that the discrepancy was largely due to massive vote-switching fraud?

Oh, and details of how the incoming vote-returns are incorporated into the exit polls projections, are given on the E-M website FAQ, which was available before the election, and is the way exit poll projections have been made for years. So let's lay that conspiracy theory to rest too.

[anaxarchos]

66. You apparently "stopped reading as well"...

You are always free to argue with yourself, of course.

But you are not talking to me.

Let me repeat again, the exit polls themselves DO NOT support your theory of an increased Kerry defection rate hidden by response error on the retrospective question. They suggest the opposite. Thus the retrospective question in the exit polls is rendered moot, regardless of this response error "debate".

Just like TIA said...

[Febble]

71. Agreed that I was making a tangential point

and probably made it badly.

I don't recognise the theory you ascribe to me though, although I may simply misunderstand your paraphrase. All I am hypothesising is that Kerry voters either had a slightly higher response rate than Bush voters in a majority of precincts, or were selected at a slightly higher rate (I think the evidence is more in favour of the latter). And that once the responses were reweighted to correct for this, the cross-tabs indicated that a minority of respondents had voted for Gore.

This would damage my hypothesis (for which, I would argue, there is considerable evidence) if I couldn't propose a hypothesis to account for it. But I can, which is that it may reflect the fact that of those who voted for Bush this time, some were Gore voters who thought they'd voted for Bush last time.

There is clearly non-communication going on here, because I don't really understand why this is not fairly obvious. I don't mean obvious that it did happen, just obvious that it could. I certainly don't buy the assertion that no Gore voter who voted for Bush would forget they'd voted for Gore. It seems to me perfectly possible that they could, and indeed it seems evident, from OTOH's data, that some did.

[anaxarchos]

73. This just isn't that hard.

I ascribe nothing to you....

I am responding.

OTOH: Bush could have won the election if Gore voters from 2000 had defected to Bush in 2004 at twice the rate that Bush 2000 voters defected to Kerry in 2004.

Several DU'ers: That is not what the exit polls say. The exit polls suggest a similar rate of defections for Kerry and Bush from voters who voted for Gore and Bush in 2000. Take a look at the "Who did you vote for in 2000?" question. Given the turnout and the break toward Kerry of new voters, Bush could not have legitimately won the election.

OTOH: That question you rest your case on is subject to response error. Gore voters could have simultaneously defected at twice that rate AND it would not show up because they would characterize themselves as Bush voters in 2000 despite the fact that they actually voted for Gore.

Me: Let's forget about the "Who did you vote for?" question. We can determine defection rates independently of that question which is, itself, not "context sensitive", etc. (i.e. the response error is an artifact of the question itself and independent of the other survey questions). Let us look at the party ID, "ideology", the GOTV questions, etc. (there are others, BTW). Do they support the notion of a differing internal defection rate between Gore and Bush voters from 2000 or from 2000 to 2004? No, they do not. Thus...

Two Conclusions:

1) The small point - There is no support in the remainder of the survey questions for the response error thesis that OTOH has advanced (this is not a generic refutation... merely that the answer to the "How did you vote?" question jives with the remaining survey questions and thus does not indicate "his" response error in this poll).

2) The MUCH larger point - There is no support in the Exit Polls for a "Gore 2000 voters defection" thesis. Quite the contrary, they indicate a re-run of 2000 (the conventional wisdom) with a much greater turnout. But, if we rerun 2000 in 2004 with a larger turnout, Kerry wins.

Thus, fraud...

Quite Easily Done (QED).

None of this is impacted by anything that you may think or believe. On the contrary, since we have just calculated internal defection rates (or "affirmed" them), and the political environment we started from is a given (the 2000 elections), we have just come up with a reason to suspect "fraud" EVEN IF the 2004 Exit Polls had not indicated a Kerry victory (which they did). rBr or not, the basic internal arithmetic does not "add up".

Ya see?

It really is "The Return of the Clincher"...

[foo_bar]

74. so the clincher is, there is no clincher?

There is no support in the Exit Polls for a "Gore 2000 voters defection" thesis.

There's no Exit Poll support (in caps, like Divine Truth) for plenty of things that actually happen. Take time and space, for instance:

From our post-election survey, the exit poll interviewers reported the distances that they were forced to stand from the polling location on election day:

Location of Interviewer # of polling locations mean WPE * Miss Rate
Inside the Building 506 38% -5.3 9%
Right outside the entrance 235 17% -6.4 10
10-25 feet away 239 18% -5.6 11
25-50 feet away 165 12% -7.6 13
50-100 feet away 148 11% -9.6 16
More than 100 feet away 53 4% -12.3 18

http://www.exit-poll.net/election-night/EvaluationJan192005.pdf

And that's before the Febble function (as taught in undergrad statistics, unlike the quasi-Kabbalahist credo), so you might imagine what kind of scatter plots you get from a universe of fickle mushheads (to paraphrase Mayor Quimby). That's why Asimov's Psychohistory is still science fiction, and not something TIA worked out in Excel.

Thus, fraud... Quite Easily Done (QED).

That's axiomatic without inserting the false dilemma between faith-based tabulations and faith-based polls. Both can be wrong without contradiction, the paradox only appears when you insist that garbage out proves flowers in. Hence the TIA clincher and the anax game are both reversals of the burden of proof, because there isn't much to build on when the premise is a set of exit polls with Bible Code properties*.

"Has it come down to this? Laymen debating experts about technical matters in which they are totally ignorant?"
-TruthIsAll, before the election
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=102&topic_id=60466#60547

*

When the authors used a randomization test to see how rarely the patterns they found might arise by chance alone they obtained a highly significant result, with the probability p=0.000016. Our referees were baffled: their prior beliefs made them think the Book of Genesis could not possibly contain meaningful references to modern-day individuals, yet when the authors carried out additional analyses and checks the effect persisted.

That is, the probability of getting the results they did was 16 out of one million or 1 out of 62,500. The authors state: "Randomization analysis shows that the effect is significant at the level of 0.00002 the proximity of ELS's with related meanings in the Book of Genesis is not due to chance." Harold Gans, a former cryptologist at the US Defense Department, replicated the work of the Israeli team and agreed with their conclusion. Witztum later claimed that, according to one measure, the probability of getting these results by chance is 1 in 4 million. He has apparently changed his mind and now claims that the probability p = 0.00000019 (1 out of 5.3 million).

http://skepdic.com/bibcode.html

[anaxarchos]

77. Can't talk to you f_b...

...even though I want to.

This kinda talk creeps me out...

[foo_bar]

81. If you can't tell me, can you tell Mr. Binky?

*gestures with hand puppet*

[Febble]

75. OK

You wrote:

Let me repeat again, the exit polls themselves DO NOT support your theory of an increased Kerry defection rate hidden by response error on the retrospective question. so I assumed you ascribed the theory to me. But whatever.

The fact is that I don't have a theory of how Bush could have won. I don't think we can tell, and I certainly don't think we can tell from the crosstabs. Not only do I regard the 2000 vote question as unreliable (and, moreover as unreliable in the direction that would give them some plausibility, cf OTOH's comment about 2000) but we know they were subjected to metamorphic pressure by the reweighting process, which was unusually heavy. It wouldn't surprise me if some stuff that was not held constant (like gender) got a bit warped.

And for those reasons I certainly wouldn't use them to support a conclusion that the exit poll discrepancy must have been due to fraud, in the presence of evidence that I do put considerable weight on that it wasn't (but which I won't rehearse here).

Thanks for the explanation. I think I understand your point better.

But no clinch.

[anaxarchos]

78. Well, let me help some more...

...before I answer "the other guy". (Yawn) ... it might end up being tomorrow.

There is no question that the survey questions were subjected to "metamorphic pressure" and some were "warped". That is precisely why they can be relied on. They can be relied on to be something between the actual survey results and a skew toward Bush resulting from the reconciliation of the Exit Poll with the vote count. With regard to internal arithmetic, they can be relied on pretty heavily. Presumably, "the metamorphosis" changes "how many", but not "of what" (or "whom").

This is the most accurate survey we have, EVEN IF Mitofsky was right (which he isn't).


Is too clinch...

[Febble]

80. Is not clinch.....

OK.

We have, broadly, two hypotheses to account for the differences between the two sets of cross-tabs:

1. the crosstabs based, apparently, on data not reweighted in light of the vote returns


2. the cross-tabs "forced" to the incoming vote returns

Hypothesis A (crudely stated): There was no bias in the poll, therefore the cross-tabs are good, and show Kerry correctly winning, with a plausible demographic story to back it up. In this narrative, the discrepancy between the poll proportions and the vote-count proportions arises because of massive fraud in the vote-count.

Hypothesis B (also crudely stated): There was bias in the poll. In this narrative, the discrepancy between the poll proportions and the vote-count proportions arises because Bush voters were under-represented in the poll.

I don't need to argue the case for A (you and TIA have done so eloquently), so I will argue the case for B.

1. There is evidence for bias in the poll.


1. a number of states had large exit poll discrepancies, and yet the vote-count was in line with pre-election expectations, or, in one case, with an apparently clean recount (NH)


2. "redshift" tended to be greater in precincts where random sampling protocol was more likely to be compromised


2. there is evidence that fraud was not the reason for the discrepancy


1. "redshift" is not correlated with "swing" in either Ohio (ESI) or nationally (Mitofsky)


2. "redshift" in exit polls is not correlated with "redshift" in pre-election polls.

OK - IF we accept hypothesis that the discrepancy was due to bias in the poll (I am not asking you to accept it, just asking you to follow the logic), then we postulate that for some reason or another Bush voters were under-represented in the poll.

What we don't know is which Bush voters. They could have been a representative proportion - i.e. the Bush voters who should have participated but didn't could have been drawn from the same population as Bush voters who did. But there is no way of knowing this, and given that we are postulating that the overall sample of voters who did participate was not drawn from the same population as those who should have participated but didn't, there is no particular reason to suppose that the Bush voter populations were the same either.

In other words, just as we are postulating that Kerry voters were over-represented relative to Bush voters, we could postulate that some kinds of Bush voters were over-represented relative to others. For example, it is possible that the "missing" Bush voters were disproportionately Gore-Bush defectors. If so, then the reason that the un-reweighted crosstabs apparently got the Gore-Bush proportions "right" may have been that Gore-Kerry voters were over-represented in the poll, but Gore-Bush voters under-represented. Therefore, when the reweighting was done, the Gore-Bush proportions got squished. Not, on this reading, because of misreporting, but simply because the Gore-Bush defectors weren't in the sample.

This is what I mean when I talk about the metamorphic pressures. Once we hypothesise bias in the poll, we don't know who to assume is missing. The reweighting algorithm will assume age, race and gender proportions are as indicated by the non-response data on those characteristics, because they are available. But the reweighting algorithm will have no data on invisible characteristics.

So no, it isn't a clinch. It's a lemon. You either buy the fraud explanation for the discrepancy, assume the poll was unbiased, and interpret the cross-tabs as a fairly good source of info as to who voted for whom. Or you buy the biased poll explanation for the discrepancy, and treat the cross-tabs with the utmost caution because you are assuming the poll was biased, but you don't, and can't, know precisely how.

And then there's the well-attested phenomenon of misreported past vote to deal with on top of that.

So I simply don't accept your statements: "There is no question that the survey questions were subjected to "metamorphic pressure" and some were "warped". That is precisely why they can be relied on." The two statements sound paradoxical, and that's because I think they are. I think it would be more logical simply to regard the un-reweighted cross-tabs as correct. The trouble is that that requires one to accept the likelihood of massive vote-switching fraud, which I think, after due consideration, is contra-indicated by other evidence.

And no, I'm not desperate, just open minded. It's what a scientist should try to be.

[Febble]

82. Shorter Febble:

Presumably, "the metamorphosis" changes "how many", but not "of what" (or "whom")

Right.

It won't change:

a White Middle-Aged Male reportedly-Bush 2000-voting Republican who voted for Kerry, into
a White Middle-Aged Female reportedly-Gore-voting Democrat who voted for Bush.

But it might apply equal upweights to:

a White Middle-aged Male reportedly-Gore-voting Republican who voted for Bush (WMMGRB), as to
a White Middle-aged Male reportedly-Bush 2000-voting Republican who voted for Bush (WMMBRB)

because the re-weighting algorithm doesn't have access to data on whether these two groups of voters were equally under-represented or not (because there isn't any). And if they weren't, it won't get the re-weighted crosstabs right, even if it gets the 2004 vote proportions right, and the age, race, sex characteristics right, even if the voters correctly reported their past vote. If, for example, WMMGRB voters were more under-represented in the poll than WMMBRB voters, then the weight would be too low for the first group, and result in an underestimate of Gore voters in the reweighted cross-tabs.

Well, not a lot shorter, but I hope clearer. And on top of that, there's the mis-reported vote thang.

As I said, a lemon, not a clinch.

(edited for typo)

[OnTheOtherHand]

76. do you really think that false reports won't affect the defection rates?

This is pretty cool. The clincher started out by embracing the "2000 vote" question, and now you are saying, "Let's forget about" it. Only not really, because you still want to lean on it to buttress the notion that the Gore->Bush and Bush->Kerry defection rates were pretty similar. And then, as long as I can't use party ID and ideology and GOTV and who knows what else to prove anything about Gore and Bush voters -- which I should think would pretty much stand to reason, since the motivated Democratic base is hardly the portion of Gore voters whom one would expect to defect, in 2004 or retrospectively -- well, then, the defection rates are probably pretty similar. That thing in the NES panel about false reporting of the past vote being correlated with defection was probably some strange artifact of the NES.

You might try that with the 2000 exit poll. The Clinton->Bush defection rate is about twice the Dole->Gore defection rate, which seems fine, maybe, so who is to say that the overreporting of the '96 Clinton vote influenced the defection rates? Let's just see how this works out. OK, for the sake of argument, let's assume that 1996 Did Not Votes really were 13% of the 2000 electorate, and then allocate the other 87% proportional to official 1996 vote shares (so Clinton's 49.2% and Dole's 40.7% become about 43.6% and 36.0%, respectively). And let's take the row percentages (defection rates and all that) and apply them to this actual electorate. Suddenly, Bush wins the 2000 election by about five percentage points. Ouch.

So, whaddya think? Does false reporting of past vote really influence the defection rates? I'm thinking yes. And I think your abrupt appeal to lots of other tabs is -- however you may have subjectively intended it -- in practice a desperate attempt to change the subject, to flee the logic that made 43/37 interesting in the first place. All in the service of your master narrative, which is that all the Gore voters from 2000 were angry, and turned out and battled all the Bush 2000 voters to a standstill, whereupon the new voters pulled Kerry over the line. That kinda makes sense if you focus on partisans, because weren't Bush's late-term approval ratings the most polarized in history? But the polarized partisans generally aren't the ones who forget whom they voted for.

(EDIT: "marginal" malaprop)

[anaxarchos]

79. You may want to reformulate this...

...before you invite me to respond to it. This one is really a stretch.

First, response error is response error is response error. It don't correlate to nuthin', or more properly, it changes from survey to survey, year to year, etc. Even the better understood "Did you vote?" isn't that well understood. "It ranges from a little to a lot...", etc. (12% to 18%? 10% to 20%? 13% to 22%?). Thank you very much. Gimme a break.

Second, TIA pointed to the 43/37 split as evidence that the "metamorphic pressure", as Febble calls it, needed to reconcile the actual vote to the exit polls was not possible, i.e. that the exit polls were not reconcilable with the actual vote. The actual exit poll numbers, prior to "blending", were around 41/39 if I remember correctly. They had to be put under some impossible "metamorphic pressure" in order to turn this lump of tar into a diamond. I'm the one that added the next piece, that this was because of the internal math of Bush/Gore returning voters from 2000. You then argued that a 43/37 split was possible because it could be caused by response error to the "who did you vote for in 2000?", question. The inference was that response error made that question unreliable (or if you REALLY push it, "reliable" in its inverse) and masked a greater defection rate, Gore->Bush, than might otherwise be indicated. This has now become your theory, repeated here, as to how Bush won.

How does one challenge that except by reference to other tabs? We look at the other tabs to see if your argument is sustainable. Is it? Nope.

BTW, the reason for going to several other questions was to take differing swipes at the defection rate (and to increase proportionately the percentage of defectors who would have to exist in the remaining "un-sampled" portion). By the time we hit Florida, we were sampling virtually all Bush/Gore voters from 2000 (perhaps 70% of a voting population which had increased by 40%).

My logic stands. You will have to take a different whack at it. Ask your psychologist friend who is sounding "desperate" by this point.

As far your 2000 scenario goes, I suppose we have to start with inferring what part of each elections' sample actually voted in the previous election and how the response error differed year to year, and .... No thanks. What a sink hole.

And no, I do not expect to move you or to be taken "seriously", etc. I have long ago decided that you are a dedicated partisan of your cause.

I'm just polishing off some of those smug edges...

[Febble]

83. "blending"

isn't really the word. The word you want is weighting. Each response is weighted. Initial weights are by age race and sex, to compensate for discrepancies between the age race and sex characteristics of non-respondents versus respondents.
Other things, including pre-election polls, may also affect initial weighting, but I don't know the details.

But as the vote-count data comes on line, this is also used as input into the weights.

We can actually see the weights in the archived National dataset: ftp://ftp.icpsr.umich.edu/pub/FastTrack/General_Election_Exit_Polls2004/Datasets/_National/

in a column marked "weight". Each row is a response from an individual voter, and each voter has a "weight". And if you use the weights you get the weighted result, which, as we know, matches the vote count. In fact you can do your own crosstabs, although it only gives the final weights.

[anaxarchos]

85. Thank you Febble, for your help...

If you notice, I have been playing with that word: "blending", "reconciling", "weighting", "corrupting", "metamorphic pressure" (yours, which I like very much... I'm just wondering whether the "medium is the message"...

Thanks always, for the help.

[OnTheOtherHand]

84. oh, golly

"response error is response error is response error. It don't correlate to nuthin', or more properly, it changes from survey to survey, year to year, etc."

Uh, gee, I would have thought that would more likely be my line, only I would use "non-response error" instead -- whereupon I would get clobbered, not altogether inappropriately, for my "nihilism." If you duck down that rabbit hole, then I think it takes some nerve to assert that the exit polls indicate anything whatsoever.

Last I checked, you were staring at a tabulation that indicated that the Democratic share of the electorate was two points smaller in 2004 than in 2000 (37% versus 39%), and that the Democratic defection rate to Bush in 2004 was about twice as large as the Republican defection rate to Kerry (11% versus 6%). And apparently the one thing you regarded as reliable in the tabulation is that at least the Democratic defection rate isn't higher than it was in 2000. (And you apparently take this point as strong support for your view about Bush2K and Gore2K defection rates, although many Gore2K voters weren't Democrats, and many Gore2K voters may no longer be Democrats, and Bush2K voters don't even appear in this analysis.) I could hardly have imagined a more contrived argument. And you call me a dedicated partisan? Goodness. I don't claim to know what the real numbers are, but at least I don't pretend that I can arbitrarily let some vary, hold others constant, and call the result a "clincher."

You can't really go back to the 2004 "voted in 2000" tab, because your premise that the defection rates there are reliable blows up, very messily, if extended to the 2000 "voted in 1996" tab. So you blow some smoke about response error and turnout and sinkholes, and hope that no one notices. Hey, let's not even consider how people say they voted four years ago. Whether they say they were contacted by the Kerry campaign is probably a better indicator anyway. Yeah, that's the ticket! Who but a dedicated partisan could deny that the GOTV numbers are a slam-dunk for equal defection rates? Who but a dedicated partisan could take the NES panel seriously when it yields a larger Gore->Bush defection rate? Hey, I probably made those numbers up in support of my cause. Wow.

AFAICT, your principled position is to take the exit polls seriously -- even literally, but always selectively -- when you think they support your prior, and shrug "who can say?" when you don't. (I believe the term of art for this exegetical principle is "fundamentalism.") Here we have had an unusual opportunity to watch you apply this principle to the very same table in opposite ways -- unless you proposed to "let's forget" the recalled 2000 vote question back in August, and I somehow missed it. This is, to my mind, the clearest illustration of "metamorphic pressure" in the entire thread.

The true believers set out to pry open Mitofsky's Black Box, citing purported evidence of Irreducible Complicity. Now, at best, the evidence amounts to one big muddle. But there will always be age estimates to question, there will always be gaps in the fossil record, and these will always point unambiguously to what the true believers already knew. They can always demand that someone explain exactly how everything happened, and when no one can, they can claim it as proof that the scientific consensus is a sham. It will be lonely -- they will be ignored by most, criticized by many of the rest -- but at least they will never risk refutation.

[anaxarchos]

86. Now you're just insulted...

Would you like a short timeout before we continue?

[autorank]

54. Anaxarchos, Great Post. People need to actually work for a candidate to

understand the energy level of a campaign. There's a concept in business accounting called "the reasonableness test." If numbers look goofy on their face, you say they don't pass the reasonableness test and tell the finance people, "check this out, now!"

Well, we've been checking things out for a while and there is no reasonable way to come up with a Bush victory (aside from the obvious fact that those who disagree continue to avoid--the Republican 'culture of corruption' will grasp any opportunity to steal power, including death and destruction, to wit Iraq).

Here is the quotation I like:

"For the survey question, “Were you contacted by the Kerry campaign?”, 26% answered yes, of whom 66% reported voting for Kerry. To the matching question, “Were you contacted by the Bush campaign?”, 24% answered yes, of whom 62% reported voting for Bush."

This shows equivalence, which takes away the only viable alternative "behavioral" explanation for a Republican win. The fundies are out of it, the youth vote canard fell quickly, and you've exposed the other deception offered by the Republican "masters of deception."

btw, Maybe you can clear this up for me. When people talk about "false recall" are they talking about voters forgetting who they voted for between the voting booth and the exit poll interviewer? If this is the case, aside from the obvious absurdities, one cannot maintain the "false recall" position and support parallel elections? N'est pas?

Thanks for contributing to this thread.

Here are Full_Metal_Hat's brilliant words from the 123 vote Greatest thread:
http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=132x2354129

"This anniversary remembrance is here to help us celebrate the voice that poured out through us, through DU, through the country ultimately. A voice revolted by knowledge of deep corruptions and even deeper cover ups. A voice ultimately muffled by a media overtaken by its own infestation. A year later we have more than enough proof of the complicity of the press, and the very act of not reporting the events of January 6th can now quite readily be seen as the proof that there was guilt to be hidden, that there were crimes being covered up. The only event that has the actual power of choosing the president and the administration was relegated lines like "Two members of congress raised a question..." when every journalist knew there were four of the most HEATED hours of debate EVER in Congress. Shame on them all. "

[Febble]

55. I can answer one of those

No, "false recall" in this context does not mean "voters forgetting who they voted for between the voting booth and the exit poll interviewer".

One advantage of exit polls over pre-election polls is that a)you know that the person did actually vote and b) you can be fairly confident they will remember who they voted for.

In this instance, "false recall" refers to the well-attested phenomenon by which people forget who they voted for at the previous election. In the dataset OTOH cites, the same people were asked who they voted for shortly after the 2000 election, and again four years later. So their original answers are on file. And several years later, it appears that they forgot (although the interviewers knew) what they'd said back in 2000. And it appears that a proportion of those who'd said, back in 2000 that they'd voted for Gore, by 2004 thought they'd voted for Bush.

Now, as OTOH says, these people may not be typical of the sample interviewed by exit pollsters. Indeed they are likely to be slightly more politically aware, as they know they are part of the panel, which if anything might make them less likely to forget. But forget they nonetheless do. And they do it over here in the UK too. Not always in the direction of favoring the incumbent government. A poll conducted a couple of years after John Major narrowly defeated Neil Kinnock's Labour Party showed that Labour had actually "won" a landslide. Why? Because people tend to align their recall with their current views, not maliciously I'm sure. But it's not easy to remember what you thought now you no longer think it.

And it appears, from the study OTOH cites, that a small but significant proportion of Gore voters had, by 2004, forgotten that they had not voted for Bush.

So it would be at least surprising if this phenomenon did not crop up in the exit poll as well.

If you want to believe that BushCo stole the election, fair enough. I think telling people that Saddam had WMD's and was implicated in 9/11 was a great way to steal an election, and that there were other foul methods as well.

But I'm not going to believe that vote-switching fraud was responsible for the exit poll discrepancy when the evidence is telling me it probably wasn't. To insist that it is is to risk fooling yourselves into thinking the answer to a Democratic victory is some kind of fidget with the electoral machinery, when the elephants stomping around the room are systematic racism in the way elections are conducted, campaign funding rules that allow the best funded campaign to buy the levers of public opinion, and the use of those levers to propagate untruths.

If you think hand-counted ballots will solve the problem, I've got a bridge to sell you (again). Not that I've anything against hand counted ballots. I'll try and sell you those too.

[OnTheOtherHand]

57. nitpick clarification

I doubt that the interviewers in 2004 knew whom people said they had voted for back in 2000! The investigators knew, in the sense that they had the data stashed away.

Also an elaboration: not only are people who participate in NES surveys -- never mind panels -- probably sensitized to be more politically aware, but we know that attrition is often higher among people with less political knowledge and awareness. For instance, Traugott and Rosenstone report that in the 1990-92 panel, about 28% of the folks who scored lowest on the political information scale dropped out, compared to 6% of the folks who scored highest. Of course many of these people don't vote anyway.

[Febble]

58. Sorry, my bad

I should have said that. Sure, if it wasn't single blind it wouldn't have been valid.

[OnTheOtherHand]

56. the clincher is dead -- long live the clincher!

We seem to have moved on right sharp. Forget about that Gore/Bush thing, it was so 2005. Now we're going to examine the approximately 35% of respondents who indicated that they had been contacted by the Bush and/or Kerry campaigns, and infer that they demonstrate "equivalence." (Note to anax: within this group, about 38% indicated that they had been contacted by both campaigns, which makes one wonder -- on your interpretation of the question -- whether the parties are so proficient at identifying their base after all. (Edited to clarify that there was only one GOTV question.))

Sorry, folks, but I just don't think that particular clincher is going to work for you. Care to try again?

And then there was that odd bit where anax apparently changed "Gore voter" to "Democrat" and hoped no one would notice. Where did I claim that the defection rate was higher among Democrats in 2004 than in 2000? Why would I have to claim this? And how does this line of reasoning lead to the conclusion that Kerry won the popular vote? (And WTF do suburban soccer moms have to do with any of it? Talk about red herrings.)

If anyone actually has questions about all this, please feel free to direct them to me. Mere reason avails little in the face of self-referential certainty.

[anaxarchos]

63. The issue is far more debatable than some around here would like...

"False memory" in this context is something of a misnomer. There are a few survey questions which trigger a certain kind of response error. The famous one is "how often do you go to church?". Another one is "did you vote?". There is a "nature/nurture" type of debate on these as to whether this is a kind of (convenient) memory lapse or whether it is a direct reaction (a "lie") to a percieved social pressure to vote or to go church. From an experimental standpoint, researchers have reported more success in addressing this as "social pressure" rather than as "false memory", though the results are not definitive.

A variation on this is the "who did you vote for?", question. This seems to be subject (sometimes) to a "bandwagon effect", i.e. people report voting for an incumbent in elections even if they didn't do so. The rub comes in when the effect is most pronounced among those who would have voted for a candidate but in fact did not vote. There is also evidence to suggest that people sometimes "switch" their vote to the winner even if they did vote. Further complicating the issue is the fact that, counter-intuitively, the effect is more evident immediately after an election than it is later on. On all of this, context matters, so that it is difficult to interpolate results from one poll based on experience with another, let alone between two different types of polls.

As far as I know, there is no research on how this is relevent to exit polls. Presumably, the fact that exit poll respondents have just voted would have some impact on the pressure to vote, even retrospectively.

[OnTheOtherHand]

64. reality check

Gee, who around here is in the habit of denying that issues are debatable? Honestly! Why don't you have a nice long heart-to-heart with your "tennis partner," and then we might be able to have something approaching a serious conversation about this issue?

I'm puzzled by your conviction that a bandwagon effect "is more evident immediately after an election than it is later on." I haven't seen that yet.

It does seem likely that the bandwagon effect is typically smaller in election studies than in the General Social Survey, and reasons spring to mind.

I was impressed that you managed to wander through the 2000 exit poll results without mentioning this result: according to CNN.com, 46% of (weighted) respondents reported having voted for Bill Clinton in 1996, 31% for Bob Dole, 6% for Perot, 2% for someone else, and 13% not having voted. Thus, Clinton had about a 17- or 18- point advantage among people who reported having voted. Clinton's actual winning margin in 1996 was about 8.5 points.

Hmm. Do you suppose that Gore stole millions of votes in 2000, and the exit pollsters had to upweight the 1996 Clinton voters by an impossible proportion in order to match the incredible official returns? 46% of the 2000 electorate would be about 48.5 million voters, but Clinton only got 47.4 million votes in 1996, which means there would be something under 46 million surviving Clinton voters in 2000, right?

Well, maybe several million votes were stolen from Clinton in 1996? Of course, the exit polls apparently came relatively close to the official returns in both 1996 and 2000.

BTW, do you happen to know whether cherry-pickers are unionized?

[anaxarchos]

68. Citations:

(1) Stanley Presser, "Can Context Changes Reduce Vote Overreporting?", Public Opinion Quarterly, Winter, 1990.

(2) Robert H Prisuta, "A Post Election Bandwagon Effect? Comparing National Exit Poll Data With a General Population Survey", AARP, 1992

"Presser (1990) reports that such response errors tend to range between 12% and 16%. He also reports that attempts to reduce response bias by altering item context have been unsuccessful. His study did find that such error was time-related, however, with the error tending to be larger the closer a survey was to an election."


I don't have the relevant Presser quote in front of me...

The rest of your post is not serious, which is OK... but it's not funny either.

[OnTheOtherHand]

72. no, it's very serious

If no one on your team has a plausible reason why the 2004 recalled-vote tabulation evinces massive pro-Bush fraud, but the 2000 recalled-vote tabulation does not evince pro-Gore fraud, then it would seem appropriate for someone formally to admit that the argument does not hold up. (Or if someone wants to argue that Gore did steal millions of votes in 2000, that would be interesting, too.)

Actually, I'm not sure there is a relevant Presser quote. If it's really in Presser 1990, I sure don't see it; the study was absolutely not about how close the survey was to an election. (Judging from the paraphrase, the original claim might have been that people are more likely to overstate having voted in a past election the closer they are to the next election.) In any case, the primary question I am concerned with is whether people wrongly report having voted for the incumbent, not whether they wrongly report having voted at all -- although obviously those questions are related.

[autorank]

88. False recall; phantom papers; food services menus.

I think I may have a case of false recall. I was surfing the net the other night and found a paper on a brilliant internet poster demonstrating that Kerry actually won in 2004. The entire paper (it looked like a journal article) was devoted to the poster. Then I went back to look for it again and it was gone. Is it false recall, a phantom paper, or did food services just take precedence?

Inquiring minds want to know?

[autorank]

89. Enough with the "false recall"...brings back bad memories.

Anax., help me here. Don't the exit polls indicate that Gore and Bush voters turned out in 2004 in equal proportions? If so, Kerry won. Why is memory even an issue if the polls show the same proportions showing up in 2004? This false recall business seems to be a distraction. The memory that's most relevant is the memory of voters between the ballot box and the exit poll interviewer. Certainly no one can argue against that process. That voting on election day was free and fair. It produced a 3% Kerry victory at the end of the day, shortly after midnight with the fourth NEP. It matched with the State Exit Polls. It was unencumbered by the likes of Secretary Blackwell and those dedicated to committing one of the great race crimes of this century, election 2004 in Ohio. The fifth NEP, modified to fit the corrupted vote, cannot be taken seriously. Importing questionable election returns as an input, which seems to have happened. It is like adding some Red Mountain Wine to a vintage bottle of Cabernet just to please the boss. It's possible but distasteful and somehow, it just doesn't fit.

[anaxarchos]

90. You've kinda got your finger on the problem...

The issue is that there are 2 National Exit Polls: The actual poll that was taken on election day and the one that is "adjusted" to the actual vote count. The original theory was not intended to "corrupt" anything. The idea was to "correct" any flaws in the exit poll against the actual vote so that the survey questions, etc. were still legit. The way this is done is by incorporating more votes from the winning side than might be indicated by the actual sample itself.

The problem is that this depends on the election to be legit. That's not a very good assumption this year or perhaps in several prior years. The problem is compounded because the Exit Poll guys are deathly averse to just saying: "This is what the Exit Poll said and this is the 'adjusted' exit poll which has been corrected to match the vote". I suppose the innocent explanation is that it's to avoid "controversy" (as if it avoided "controversy" in this election... sheesh).

What then happens is that people like us are left to try to discern what the Exit Poll said about the election in doubt. There really is no alternative because it is, by far, the most accurate survey available. It's taken on election day, the people interviewed actually voted, etc. It's not an opinion poll taken two weeks later, with post mortum opinions "adjusted" by Fox News, etc.

A further complication is that from our standpoint, the Exit Polls can indicate a problem (i.e. "fraud") in two entirely different ways. First, the sample itself can indicate that the loser actually won (which the 2004 Exit Polls did). That then leads to the one year debate over whether the sample or the vote was wrong ("The sample was wrong because Bush responders were reluctant to talk to exit pollsters"; "No they weren't, the numbers don't indicate that...", etc.).

Second, whether the samples were right or wrong or how they were "adjusted", presumably doesn't effect what they said "internally" (that is, what percentage of Democrats were afraid of terrorism is not radically changed by whether 1000 or 950 Democrats were counted in the adjusted poll).

TIA turned this into his "Clincher" by noticing that the adjusted Exit Polls had Bush voters from 2000 coming back in 2004 in numbers that were impossible (more of them than actually existed). This pointed to the arithmetic of the 2004 election (that the proportions reported in the survey could not support a legitimate election for Bush). OTOH then came back and argued that this question (who did you vote for in 2000?) was subject to a known error (this is where all the "false memory" talk comes in). Because this error is not easy to quantify, the next round of debate started ("Is too 'false memory'"; "Is not!").

Then, OTOH went one better by saying that Bush could have won the election through a defection of Gore voters to Bush and this known error would explain why that defection was not apparent in the exit polls. Then, I said, look at all the other survey questions that don't contain your known error. They don't support the idea of a larger defection of Gore voters to Bush than are indicated in the exit polls meaning that TIA is right. Then he got all pissed off, presumably because if I don't like the adjusted exit poll I am not allowed to use the other questions from it(I can't really tell because he is busy calling me the antichrist at the moment). To be fair to him, I also poked him (couldn't help it... he kept saying all this insulting shit).

The net net right now is as you say. We are back to the conventional wisdom here that the election given the turnout, and the proportions reported in the exit poll suggest Kerry won. Admittedly, this is a pretty tortuous road back to November 5th of last year when most people here had already come to that conclusion.

Of course that's just my opinion of the current status.

I am sure the "debate" isn't over...

[Febble]

91. No, the debate isn't over

and I think this is a valuable post.

I think the two "sides" have been talking past each other, and your post clarifies both your own argument (to me) and your perception of that of the "opposition". I had definitely misunderstood yours. I believe you have misunderstood OTOH's, and possibly mine.

It also seems clear that not only have each "side" been misunderstanding each other's arguments, but each side has felt increasingly insulted, maybe rightly, maybe not.

OK, maybe it's not my role to play umpire, but it seems to me that whatever may divide us in terms of math, or interpretation of the math, we all want to know why Kerry isn't president, and we all grieve that he isn't (because he damn well ought to have been).

So let's abandon the tennis thing, and figure out how to play on the same side.

First rule: we stick to the arguments.
Second rule: if we don't understand an argument, we ask for clarification.
Third rule: we play nice (and that includes incorporeal participants).

OK?

Maybe a clean thread would be nice, back home. After a timeout, I'll start one.

[anaxarchos]

93. You're right, Febble...

It would be inappropriate for you to play referee. It would be much like one of the opposing defensive ends suddenly grabbing the whistle, in the fourth quarter yet, and declaring that we'll just play this over someplace else (and to my new, less violent, rules to boot...).

No thanks. It was hard enough to get to here (and my ribs are a little sore).

Besides, as we have discussed, I don't believe we are on the same side (at least on this issue). Awsi Dooger doesn't think so either. He is gamely running on to the field at the eleventh hour to try to save his team (he probably has money down). Admittedly, he's wearing hockey skates and carrying a basketball, but hey... he's trying.

Finally, I haven't debated this issue since TIA got banned, in mid-paragraph, while he was pointing to this very subject. It didn't seem fair to me to rehash this stuff without him. Somehow, on this thread, I felt his "spirit".

If we are done (forever, for now, until the "next round"... whatever), that's OK.

It works for me....

[Febble]

94. Works for me too

I was in danger of getting interested in those crosstabs, and I've got a dissertation to finish....

See you whenever.

Cheers

[autorank]

59. Anax: I was just "Channel Surfing TIA" and voila...

The Final NEP 43/37 weighting is an eternal albatross which
will never explain without resorting to the mantra:
False recall, false recall, false recall.
Forgetful Gore voters, forgetful Gore voters, forgetful Gore
voters.

It's catapulting the propaganda.
It's classic obfuscation....

...of how Gore voter "false recall" explains their 
implausible hypothesis that 14.6% voted for Bush.

Well, I don't see it, not because the logic is quite
convoluted, which it is, but because I don't accept the
premise.

It doesn't pass the smell test.
It's a Hail Mary pass.
It's designed for Immaculate Deception.

In a bizzaro universe, 14.6% of Gore voters switch to Bush. 
In a rational one, Gore voters were out in force for Kerry -
to kick the thieves out of the WH.

In a bizzaro universe, only 52.6% of DNV voted for Kerry.
In a rational one,  massive Democratic registrations brought  
new voters out in droves (55-60%) for Kerry.

In a bizarro universe, a 48.5% average Bush approval 
rating (1% MoE)  morphs into a 3 million vote mandate.
In a rational one, a rating that low means Bush is toast. 

The related regression scatter chart of incumbent approval vs.
vote count is meant to mislead, not illuminate. It appears
interesting, until we apply full perspective to the actual
numbers:

SIX OF THE NINE ELECTIONS CITES WERE LANDSLIDES!

REGARDLESS of how high an incumbent's approval rating,
there is a REALISTIC 60% UPPER LIMIT on his final vote share.
This is a HISTORICAL FACT.

The CONVERSE is ALSO TRUE:
REGARDLESS of how LOW an incumbent's approval rating,
there is a REALISTIC 40% LOWER LIMIT on his final vote share.
This is a HISTORICAL FACT.

INCUMBENTS LOST FOUR OF THE NINE ELECTIONS CITED. 
Ford, Carter, Bush 92...AND Bush 2004.
EACH HAD AN APPROVAL RATING UNDER 50%.
OF THE FIVE WINNERS, CLINTON HAD THE LOWEST APPROVAL
RATING:55% 

BOTTOM LINE: 
THE NUMBERS PROVE THE 50% RULE. 
PERIOD.

The graph is a classic straw man, a diversion from the truth.
It uses real numbers, but hides their meaning.

Skeptics have a fondness for using scatter charts in lieu of 
analyzing the numbers which make up the scatter. 

There's a reason for this. Watch close.

Here's what the numbers are REALLY telling us:
 
			Votes (mm)	Incumbent				
Year	Pres.	Appr	Incumb	Opp.	2pty		Result	Ap>50?	Match
result?
1956	Eisen.	70	35.6	25	58.7%		Won	Yes	yes
1964	Johnson	75	43.1	27.2	61.3%		Won	Yes	yes
1972	Nixon	59	47.2	29.2	61.8%		Won	Yes	yes
1976	Ford	46	39.1	40.8	48.9%		Lost	No	yes
1980	Carter	31	36.5	43.9	45.4%		Lost	No	yes
1984	Reagan	60	54.5	37.6	59.2%		Won	Yes	yes
1992	Bush	30	39.1	44.9	46.5%		Lost	No	yes
1996	Clinton	55	47.4	39.2	54.7%		Won	Yes	yes
									
2004	Bush	48	62	59	51.2%		Won	No	No

In the three elections prior to 2004 in which incumbents were 
defeated, the average incumbent 2-party vote share was 47%.
The 12:22am NEP gives a Bush 2-party vote of 48%.

Of the nine elections, 2004 is the only one in which the 50%
approval 
rule did NOT hold up. It sticks out like a sore thumb, doesn't
it?

Query mas?

In a bizarro universe, one radically adjusts every NEP vote
share in order to conjure up an implausible Bush win scenario.
 In a rational one, a case is built based on the weight of all
 the evidence - without using tortured hypotheticals.

In coming up with the implausible Bush win scenario, sly
sleight of  hand was used to mislead readers unfamiliar with
NEP demographic calculations with the assumption that 26.5mm
2004 voters were DNV 2000 (21.7%). 

The 12:22am NEP demographic had it at 21.5mm (17.6%), so the
discrepancy is far in excess of the MoE. Increasing the DNV
group by 5mm and lowering Kerry's vote share to 52.6% reduced
Kerry's DNV margin from 4mm to 1.7mm. 

In the NEP timeline: 
Kerry's DNV share was 59% at 7:33pm, 57% at 12:22am, 54% at
1:25pm.

This is the conundrum that skeptics have to deal with: 
They cannot use the impossible final NEP 43/37 weights.
Their only "wiggle" room is to inflate Bush's vote
shares.
And they must inflate them way beyond the MoE to match the
vote count.

a) NAYSAYERS NOW USE IMPLAUSIBLE BUSH VOTE SHARES TO MATCH THE
VOTE COUNT.
b) THE FINAL 1:25PM NEP USED IMPOSSIBLE WEIGHTINGS TO MATCH
THE COUNT.

WHAT DOES THIS TELL US ABOUT THE VOTE COUNT? 
IS IT 
a) JUST EXTREMELY IMPLAUSIBLE OR  
b) ABSOLUTELY IMPOSSIBLE?

THAT'S THE QUESTION.
PICK ONE.
THINK.

[Febble]

60. Ha!

TIA: 

"Skeptics have a fondness for using scatter charts in
lieu of analyzing the numbers which make up the scatter."

Right.  Here we have the problem.  

Scatter-plots show you the actual data.  The regression line
through the scatter shows you the fit of your model to the
data; the R squared tells you how good the fit is.  The MoE of
the regression line will also tell how good the fit is.
Regression lines are the way we draw inferences from data.
Whether our hypothesis is linear or non-linear, the fit of the
model (the regression line) to the data will tell us how well
the hypothesis describes the data.  It can, if we have a
random sample, allow us to generalize from our sample to a
population.  It is how we derive "rules" such as the
"incumbent rule".  In other words it is the basis of
inferential statistics.

And it is the basis of the General Linear Model, which can in
fact, also be used to test non-linear hypotheses. It allows is
to determine the pattern of covariance between the variables
we are interested in, for example the covariance between an
incumbent's approval rating and his vote-margin.

If TIA doesn't like scatter-plots, he is not a serious data
analyst. If he wants to demonstrate that he is, I suggest that
the next time he issues a plot (column or scatter) with the
equation for the best fit line that Excel so conveniently
provides for him, that he also gives the MoE for each of the
regression coefficients in his equation.  Or an F test of the
model fit.

Otherwise I will have to conclude that he does not understand
the basis of inferential statistics.

The rest of the post appears to be summarizable as: I don't
believe Bush could have won, and I don't believe Bush voters
would have forgotten they voted for Gore, therefore the exit
polls must have been correct.

This is not statistics: this a credo.  

Which I can respect as a credo but not as statistics.

[Febble]

69. I have the spookiest sense

that somewhere in meta-cyber-space, I just trod on somebody's
toe.

If I misunderstood a meaning somewhere, I apologise.

But let me clarify a few principles regarding my own approach
to data analysis:

1. I am more interested in hypothesis testing than data
mining.  If I have a theory about what caused something else,
I devise a hypthesis that can be tested statistically. 
Designing appropriate hypotheses can be difficult, and
sometimes requires imagination.  Often they come to me in the
night.  But the important thing is that, if possible, the
hypothesis should precede the analysis, i.e. be "a
priori".  Otherwise we need to use more stringent
criteria ("post hoc" tests for
"significance" and thus risk losing statistical
power (or being misleading).

2. While I think it is impossible to be completely objective
in data analysis, two guards against subjectivity are the a
priori hypothesis and the two-tailed test. Much as I deplore
the fact that Bush is your president, and dearly as I yearned
for President Kerry, in all the analyses I have done, whether
on my own, in collaboration with other people involved in
investigating the anomalies of the 2004 presidential election,
or for Warren Mitofsky, I have applied these principles as
stringently as it has been within my power to do.  Moreover,
any finding I have made I have subjected, as I am trained to
do as a scientist, to as much probing for alternative
interpretations as I have in my power. And I remain, as any
good data analyst must, always aware that findings are always
provisional - that new findings appear that cast my current
findings into a different light, or that some aspect of the
problem that you have ignored may turn out to be crucial. 
This has happened more than once during my analyses of the
election data, and is why I like working on blogs.  

3. All statistical inferences are made with a margin of error.
 This is as true of rules like the "incumbent approval
rating" rule as it is to survey findings.  We need to be
as rigorous about our inferences in cases where they suit our
a priori hypothesis as when they do not.

3. Human beings behave in complex ways.  This mean that it
means that obtaining a random sample is difficult; it means
that generalizing from a sample to a population is difficult,
as you cannot be sure you have sampled from the population you
meant to study;  it means that the data you collect from human
beings are subject to idiosyncrasies; and it means you can get
an apparently "significant" result from one study
and fail to replicate it in the next.  There are always more
factors affecting variance in your data than you will ever
know about.

4. If you get a finding that doesn't suit your hypothesis, you
look it squarely in the face.

Well, it is my provisional view, based on serious
consideration of a fair bit of evidence, and on serious
consideration of many of TIA's analyses (I cannot promise to
have considered them all) that his contention that the exit
polls demonstrate massive fraud is probably wrong.  I don't
doubt the accuracy of the calculations, which I am sure are
less error prone than mine, but I do question the validity of
the assumptions that underlie his inferences. And I also
question, at times, his methodology. I don't think that doing
either of these things is undermining American democracy. 

In fact, if I, and others who share my conclusions, are right,
and TIA, and those who share his, are wrong - in other words
if it is the case that while the election was corrupt, massive
vote-switching did not occur, then to have people seriously
believing that the Democrats are doomed and they might as well
not vote, or that there is no point analysing Kerry's campaign
because it was perfectly OK, then I think that is a problem.

Yes, I believe that American democracy is broken and needs
fixing.  But I also believe that Kerry lost, and that needs
fixing too.

I hope that has cleared the air.  It doesn't seem to have done
anything for my html which remains stubbornly in courier. 
Weird to be posting in TIA's typescript.  That guy is
definitely spooky.

[foo_bar]

70. slightly off-topic: courier

There's a checkbox above the prospective subject line, "Check here if you want to format your message in plain text". The weird part is it comes pre-checked if you reply to a TIA seance (or any other unformatted post). Dunno if that's a bug or a feature.

[OnTheOtherHand]

62. could we harness all that spin to power a small city?

As Febble pointed out, much of TIA's post basically says, "In a rational universe, Kerry would have won, therefore Kerry won." That is the very essence of a faith-based argument, even though it uses the word "rational." Progressives should demand better.

I have no way of knowing what percentage of Gore voters (who voted in '04) switched to Bush. But dismissing the NES panel data, which put this figure in double-digits, because one Just Knows that the Gore->Bush defection rate should be no higher than the Bush->Kerry defection rate is a triumph of faith over reason. Progressives should demand better. False certainty is not what we do best.

TIA is welcome to fit an S-curve to the data and see if he can obtain what he firmly believes is the right answer. Given the firmness of his belief, I suppose he will find a way.

He will have to assume that Truman pulled his approval rating up by 10 points from the summer of 1948, but that is quite possible, so let's try what he does and start in 1956.

OK, he has three incumbents running for reelection with approval ratings under 50 (1976, 1980, 1992), and they all lose. Do three cases provide strong empirical support for a "rule"? Obviously not.

In 1976, the incumbent has a 46% approval rating and loses by 1.7 points; in 1996, the incumbent has a 55% approval rating and wins by 8.2 points. If we look at those two points, do they support the inference that the approval break-even to win reelection is 50%? No, they don't. (By linear interpolation, it would be about 47.5%.) Sure, 1996 is anomalous -- but, if we're looking at incumbents with approval ratings in the 50s, it is actually most favorable to his case. In 1972, Nixon has a 59% approval rating and wins by 18 points. Do 1972 and 1976 support the inference that the approval break-even is 50%? No, they don't. (By linear interpolation, it would be about 47.1%.)

Actually, are there any two data points in this plot that support the inference that the approval break-even for incumbents seeking reelection is 50%? I don't think so. An intellectually honest analyst might worry about that.

The only way (or the only way demonstrated so far) that TIA can get the result he requires -- because he already Just Knew the answer before he began his analysis -- is to throw out most of the information, reducing the data to "yes" and "no" (approval rating greater than 50, yes or no). Well, that is a crock. Prior to 2004, there are no incumbent-seeking-reelection approval ratings between 46 and 55. So there is no way to know whether the rule should be "incumbents with approval ratings under 50 lose, those with approval ratings over 50 win," or whether the rule should say "47," or "54."

A new rule! Incumbents with approval ratings over 47% always win! Sigh.

By cooking the data, TIA creates the impression (at least for himself) that 2004 sticks out like a sore thumb. But if one actually has the guts to look at the scatterplot -- with or without 1948, with or without all the elections where the incumbent didn't run again -- one can see that 2004 doesn't stick out.

(No wait, silly me, the graph is a diversion. It hides "meaning." Pay no attention to the man behind the curtain!)

Of course, we could always resort to common sense. Approval ratings are not exact; different firms get different approval ratings; incumbents actually run against opponents. Could we determine the "true" value of X to plug into an approval-rating rule? No. So, if I insisted that Bush's 48.5% approval rating proved that he won, that would be untenable. As TIA's approval-rating argument is, start to finish.

So, that clincher doesn't work either. Next?

(Edit to correct formatting)

[Awsi Dooger]

92. Terrific posts by Febble and OnTheOtherHand

I haven't been here in nearly a month and this is a very welcome first thread for me to sample on the rebound, after my Trojans blew the national title game. Hey, maybe TIA can twist that into a win with some creative math.

I work in sports statistics applied to sports betting. If I tried to submit a sample of a dozen or so games with wildly divergent expectations and outcomes and tried to make a definitive analysis or prediction based on that sample, I would fire myself before the boss ever had a chance. That was always the problem with 2004 and I warned TIA about that repeatedly beforehand. His projections on undecideds to the challenger and the approval rating hypothesis, etc. were always based on flimsy comparisons to literally a handful of presidential outcomes. There was never any logical tweaking. He accepted state polls without any background on how they typically erred toward actual numbers based on the partisanship of the state, or that they naturally lagged national polls. So all those flaws are inherent to his election model to begin with.

In sports betting I managed to succeed only after nailing the numerical aspect through several years of Excel research and then slowly realizing subjectivity needed to be sensibly applied. IMO, politics is identical. When a football game is at halftime with a 6 point deficit for a 7 point favorite, the numbers tell me what the historical likelihood of a comeback win are. But instead of embracing that as an absolute you need to look at applicable factors, like number of rushing attempts and yards per pass attempt by both teams, plus subjective critiques like injuries, and perhaps estimate a few percent adjustment either way, which is all the difference when the margins are so slight. In election 2004 I really don't see how Bush's win is questioned, considering the subjective adjustments all fell in his favor: incumbent with his party in office only one term; 9/11's impact on party ID toward the GOP due to perceived national security edge; don't change leaders during a war, even if it's self-made; Kerry's lack of charisma, especially considering how vital personal favorables are when trying to oust a presidential incumbent; the GOP's ability to focus on a mere handful of states.

In a polarized 50/50 era there was nothing to guarantee Kerry had enough of an edge to overcome those factors, not with Bush's poll numbers and approval ratings always only slightly below 50% heading toward election day, uncharted territory. As a bettor I can tell you flat out there was never one day all year Kerry was favored to win at any respected outfit. Bush was always between 5/7 and 1/3 chalk. I consult many of the people who work for those offshore outfits, and know how what numbers they look at to adjust their political odds. Frankly, the consensus all along among them was Bush should be a larger favorite based on the apparent security moms shift, but the public betting was keeping the number lower than it warranted. Now we're supposed to take TIA's word that it was a 99.99% Kerry absolute, the percentage he touted for weeks prior to November 2.

In '96 and '00 I printed exit poll data off the internet on election night then had to throw the originals away once they were updated. So that conspiracy is hysterical. In isolating dozens of bellwether counties before the election and looking at them in December, the slight but critical drift toward Bush in relation to 2000 was unmistakable, and that included every type of voting method and apparatus.

[OnTheOtherHand]

95. laughing about the Trojans

A few days ago I considered posting: USC wins Rose Bowl! based on the undisputed rule that in over 100 years, no team has ever lost the Rose Bowl when its quarterback passed for over 360 yards. But then I couldn't find out whether that was true -- and hey, I try to get the facts right. (Ryan Leaf came pretty close in the late 90s, I know that.)

A tip of the hat to you: I have always admired your posts, and look forward to verifying the exact meaning of "between 5/7 and 1/3 chalk."

Someone afflicted with selective literalism may tee off on your brief reference to "security moms," but I think the party ID trend is pretty striking, demographic correlates aside. And as you say, Bush did better in many, many places, regardless of voting technology.

I think Anax's #90 is pretty revealing: it was the conventional wisdom around here* after the election that of course Kerry had to have won, and that's why none of the particular arguments supporting that view has to be very compelling. It's like whack-a-mole.

*EDIT: I'm not sure where "here" is, which I suppose is why I inserted the "around."

I might try pointing out that if Democrats become Republicans, that wreaks havoc on "defection rates" if we are comparing 2000 to 2004. OK, in 2000 we have (according to the weighted and rounded tabs) 39% Dem, 35% Rep, 11% Dem defection rate, 8% Rep defection rate. In 2004 we have 37% Dem, 37% Rep, 11% Dem defection rate, 6% Rep defection rate. So does that mean that Kerry got as big a vote share from Democrats as Gore did? No, since on net 2 percentage points worth of Dems defected to the Reps. So, of the 39% of former Democrats (so that we are comparing apples to apples), 37 points defect at an 11% rate, and 2 points defect at a 93% rate. That's an average of over 15%, which is obviously substantially higher than the defection rate among Republicans. (Obviously I don't mean these numbers literally.)** So I don't see how the Constant Mean Defection Rate argument can work unless someone can prove that the 37 Dem/37 Rep split Just Can't Be Right -- as was nobly attempted, and now has crashingly failed AFAICS, with the 2000 recalled vote question.

**EDIT: In particular, I don't think that 2% of the electorate magically changed from Dem to Rep. Probably some Dems became Reps, some Reps became Dems, etc. etc., there was sampling error, there may have been weighting distortions.... Regardless, I don't see how comparing the party tabs in 2000 and 2004 supports a fraud inference.

Same general problem with the ideological comparison -- and with ideology as with party ID, there is external evidence supporting the shift. Although maybe we're not supposed to refer to external evidence these days. It is hard to keep track.

[DemBones DemBones]

96. Just a minor quibble -- TIA mentions how many voters died, but not

how many new voters there were in 2004. How many people turned 18 or otherwise voted for the first time in 2004?

[OnTheOtherHand]

97. well, according to the exits

About half the respondents in the national survey (which is a subset of the entire exit poll) were asked if they were first-time voters, and of those, about 11% said they were. If you look at that group by age, fewer than half of them (about 45%) were in the age group 18-24 -- so, maybe 5% or less of the electorate was people who were newly eligible by age.

Overall, about 17% said they hadn't voted in 2000, although I believe the true percentage was higher.

So, "new voters" can mean several different things. I don't think that counts as a quibble against TIA, since his argument doesn't depend on new voters however defined. (I disagree with him, but not on that basis.)

[TheGunslinger]

101. No kidding. That's a GLARING flaw in the logic in the OP.

Completely ignores voters who passed 18 after 2000 or were not voters in 2000 but came out in 2004 (remember all of those states with gay marriage bans on the ballot?)

[OnTheOtherHand]

105. ah, gee, you are making me pity TIA -- his arg isn't bad in THAT way

The exit polls (weighted or unweighted) indicate that the people who did not vote in 2000, but did vote in 2004, somewhat favored Kerry. Now, a lot of people who didn't vote probably say they did. And, as you say, a lot of people who didn't vote in 2000 actually voted for Bush, so we can't just say to ourselves, "Ah, new voters were young voters and, heck, they probably broke for Kerry 60/40." But TIA isn't completely ignoring the new voters. He is saying that if the new voters favored Kerry (doesn't matter so much what the margin was), then the only way Bush could win would be to do better among the returning voters. He is convinced the exits -- properly interpreted -- pretty much prove that Bush didn't do better among returning voters. He is wrong, because he is cherry-picking (unintentionally, I think, although it would be nice if he had some capacity to learn).

Sorry to nitpick, but since no one is making TIA's case very effectively, I want to try to keep the record (as I understand it) straight.

[autorank]

98. This settles it. For the record...

Bush gave us Alito, who will take away Roe, which will be
another messanger of death special delivery.  Why anyone would
defend these monsters is beyond me.  Why the monsters would
behave differently in an election than they do on other venues
is beyone me and most people.  

There is nothing to low for these people.  But the one thing
they need more than anything is to be "elected" as
per the master plan.

What a pity.

TIA:

The 12:22am National Exit Poll (NEP) (13047 respondents)
had Kerry winning by 51.41-47.62%.
On a two-party basis, the vote share is  51.91-48.09%.
This was the last pristine NEP timeline.
The Final NEP (13660 respondents) WAS MATCHED TO THE VOTE
COUNT.
AND BUSH WON BY 51-48%.

My final election model projection  had Kerry winning by
51.80-48.20%,
based on an average of 18 final pre-election national polls.
The assumption was that Kerry would win 75% of the
undecided/Nader vote. 

It turned out that my projection was too conservative. 
It was off by more than 0.11%. Here is why: 

The 12:22am NEP weighting split for the "How Voted in
2000" 
demographic was 41% Bush/39% Gore. But these weights are
impossible. 
The maximum number of Bush 2000 voters still alive to vote 
in 2000 was 48.7 million. 

Therefore the maximum Bush weighting was 39.82% (48.7/122.3)
The maximum Gore weighting was slightly higher (41.25%). 
Gore won the popular vote by 540,000. 

Using the 41.25/39.82% weights assumes a 100% turnout of 
Gore and Bush voters. Kerry wins by 52.34-46.71%, a 7mm
margin. 

Kerry's two-party vote share was 52.84% = 52.34/(52.34+46.71).
My final projection was 51.80%.

Considering the realistic, feasible  weightings, how did Bush 
win 14 million new votes compared to 2000?

Naysayers claim he won 15% of returning Gore voters, many of
whom 
FORGOT THEY VOTED FOR GORE ("false recall"). 

They claim Bush won 52.6% share of those who did not vote in
2000. 
REGISTRATION STATS SAY OTHERWISE.

These percentages are much higher than the Final NEP.
And way beyond the margin of error.

Let's calculate the probability that Bush would get those
voters.
______________________________________________________________

TruthIsAll  (1000+ posts)      Sun Aug-21-05 12:56 PM

WHAT IS THE PROBABILITY THAT BUSH WOULD GET 15-18% OF GORE
VOTERS? 
 Edited on Sun Aug-21-05 01:51 PM by TruthIsAll

There were approximately 3200 respondents to the National Exit
Poll question: How Did You Vote in 2000?			

The margin of error (MoE) for N= 3200 respondents is given by
the formula: MoE = 1.96 *sqrt(p*(1-p)/N). 

Th MoE = 1.24% for this sample. But we'll avoid using the
formula.
Otherwise, it will just divert us from the point of
discussion. 
According to Mitofsky's MoE table, the MoE is 2% for a 3200
sample-size. 

We'll use a 2% MoE  to calculate probabilities that Bush would
get: 

a) 15.5% of Gore voters assuming the Final National Exit Poll,

which Bush won by 51.11-48.48% and 

b) 18% of Gore voters assuming the 12:22am Exit poll, 
which Kerry won by 51.41-47.62%.

__________________________________________________________

National Exit Poll: 
12:22am, 13047 respondents				

Kerry wins by 4.5 million votes.

Voted	Mix	Bush	Kerry	Nader	Votes
No	17%	41%	57%	2%	20.78
Gore	39%	8%	91%	1%	47.68
Bush	41%	90%	10%	0%	50.13
Nader	3%	21%	71%	8%	3.67
98%	100%	47.62%	51.41%	0.97%	122.26
	122.26	58.22	62.85	1.19	


Using plausible weights, and assuming maximum 2000 voter
turnout,
Kerry wins by almost 7 million votes.
	
Voted	Mix	Bush	Kerry	Other	Votes
No	17.31%	41%	57%	2%	21.16
Gore	40.25%	8%	91%	1%	49.21
Bush	39.82%	90%	10%	0%	48.68
Nader	2.62%	21%	71%	8%	3.20
	100%	46.71%	52.34%	0.96%	122.26
	122.26	57.10	63.99	1.17	

_______________________________________________________________
The following  table shows that Bush needed 18% of Gore voters

to obtain his 3 million margin. 

The probability of that ocurence is less than 1 in 450
trillion,
given the 8% exit poll result.
		
Voted	Mix	Bush	Kerry	Other	Votes
No	17.31%	41%	57%	2%	21.16
Gore	40.25%	18%	81%	1%	49.21
Bush	39.82%	90%	10%	0%	48.68
Nader	2.62%	21%	71%	8%	3.20
	100%	50.73%	48.31%	0.96%	122.26
	122.26	62.02	59.07	1.17	

Prob =1-NORMDIST(X, 8%, 2%/1.96 ,TRUE)

This is a probability table for Bush to win at least x%
of  Gore voters, for X from 5-16%:

X	Prob>X	Odds: 1 in
5%	99.84%	1
6%	97.50%	1
7%	83.65%	1
8%	50.00%	2
9%	16.35%	6
10%	2.50%	40
11%	0.16%	609
12%	0.00%	22,577
13%	0.00%	2,083,900
14%	0.00%	485,887,839
15%	0.00%	288,701,537,060

16%	0.00%	450,359,962,737,050 (1 in 450 tr)


****************************************************


National Exit Poll: 1:25pm, 13660 respondents
Bush wins by 3 million votes.
But we know that 43/37 is impossible.

Voted	Mix	Bush	Kerry	Nader	Votes
No	17%	45%	54%	1%	20.78
Gore	37%	10%	90%	0%	45.24
Bush	43%	91%	9%	0%	52.57
Nader	3%	21%	71%	8%	3.67
	100%	51.11%	48.48%	0.41%	122.26
	       62.49	59.27	0.50

Changing to plausble weights and assuming maximum 2000 voter
turnout, 
Kerry wins by 3 million votes.
					
Voted	Mix	Bush	Kerry	Other	Votes
No	17.31%	45%	54%	1%	21.16
Gore	40.25%	10%	90%	0%	49.21
Bush	39.82%	91%	9%	0%	48.68
Nader	2.62%	21%	71%	8%	3.20
	100%	48.60%	51.02%	0.38%	122.26
	122.26	59.42	62.37	0.47	

___________________________________________________
Bush needed 15.5% of Gore voters for his 3 million vote
margin.

Voted	Mix	Bush	Kerry	Other	Votes
No	17.31%	45%	54%	1%	21.16
Gore	40.25%	15.5%	83.5%	1%	49.21
Bush	39.82%	91%	9%	0%	48.68
Nader	2.62%	21%	71%	8%	3.20
	100%	50.81%	48.40%	0.79%	122.26
	122.26	62.13	59.17	0.96	


To get 15.5% of Gore voters, given the 10% exit poll result,
the probability is 1 in 28 million.

Prob =1-NORMDIST(X, 10%, 2%/1.96, TRUE)

X	Prob>X	Odds: 1 in
5%	100.00%	1
6%	100.00%	1
7%	99.84%	1
8%	97.50%	1
9%	83.65%	1
10%	50.00%	2
11%	16.35%	6
12%	2.50%	40
13%	0.16%	609
14%	0.00%	22,577
15%	0.00%	2,083,900

15.5%	0.00%	28,322,049 

_______________________________________________________________


Query mas?

[OnTheOtherHand]

102. for the record, this is lame as it ever was... lame as it ever was...

No one here is "defending these monsters."  Nor has
anyone argued that the Bushies [b]wouldn't[/b] steal the
election because they are just too nice, or whatever.  You
should stop trotting out this flaming straw man.  If someone
came over and posted that Bush is a space alien, or that he
eats babies for breakfast, would it be permissible to point
out that there really isn't evidence for that?  Yes, you think
the evidence for fraud is better, and I'm happy to discuss the
evidence.

TIA is actually right that even the 41/39 Bush/Gore split in
the unweighted exit poll results is, at best, very unlikely
(although it could be due to sampling error) -- which is
another reason why he should catch on, belatedly, that what I
have said about false reporting of past vote is true and
important and not just a ploy to blow up his argument.  But I
doubt that he will.  I guess now he thinks that the exit polls
were biased by Reluctant Kerry Respondents??  I dunno, I can't
be bothered to think harder about it than he does.  His
numbers don't make sense, do they?  If the maximum possible
Bush proportion was 39.82%, and Gore had just 500K more votes,
how could the Gore proportion be 41.25%?

And by the way, he is misstating the facts (what a shock!). 
"Naysayers claim (Bush) won 15% of returning Gore
voters" -- no, only one "naysayer" (TIA's
favorite N word) has suggested a percentage, the percentage
was lower, and I've said repeatedly that I do not know and
cannot know what the percentage was.  What is needed in order
to account for Bush's victory margin is some combination of
differential turnout and differential defection between Gore2K
and Bush2K voters, and I think that differential defection
alone could do it.  TIA has never proven that it can't -- in
fact, he doesn't even argue the point.  Oh well.

[Febble]

104. Hey, you forgot the garlic!

You have to uncheck the TIA curse otherwise, the undead will haunt your formatting. Luckily it doesn't seem to give you ALLCAPs.

Well, I'm not defending monsters either, just trying to figure out how the monsters got back into government.

But here's a problem:

All TIA's statistics assume random sampling. I dispute that the sampling was random, but so far, at least TIA has been consistent.

But what's this?

Using plausible weights, and assuming maximum 2000 voter turnout, Kerry wins by almost 7 million votes.

"plausible weights"? WTF?

Why do we need weights if the sampling is random?

And if it isn't - well, why don't we all just knit our own to fit our own predictions?

You weight the Gore vote.
I weight the Bush vote -
- Let's call the whole thing off....

[OnTheOtherHand]

106. you got a weight a minute, weight a minute, oo-oo-ooo...

Yeah, it's hard to decide whether the fonts or the arguments are spookier. I thought the object was to prove that the weighted exit poll results were internally inconsistent -- but, as a matter of arithmetic, the weighted results can't really be internally inconsistent. So the Fraud Squad keeps trying to sneak in assumptions from the outside world without being noticed. Not that I object to the outside world; it's the sneaking that gets us into trouble.

[Febble]

107. Clarification on random sampling

for the sake of any unquiet spirits who may be hovering....

Voters may (or may not) have been selected at random. The protocol should have ensured random selection if it was adhered to and if fluctuations in coverage did not also coincide with clusters of like voters.

There is some evidence that selection was not random, but yes, selection was certainly intended to be random.

However, response rate is not random. It depends on voter characteristics. Voters do not toss a coin when deciding whether or not to participate in a poll. They simply decide.

The reason that age, sex and race of both respondents and non-respondents are recorded, is so that if the sample of respondents contains a different proportion of voters with these characteristics than the sample of selected non-responders, the respondent sample can be weighted to match. I have no problem with that; I believe TIA has no problem with it either.

And because this reweighting is necessary, we know that non-response bias exists. For example, it appears that men are more likely to refuse than women.

The problem is that the pollsters can only weight for visible characteristics. They know if they have a reluctant male responder problem; or a reluctant white responder problem; or a reluctant middle-aged responder problem; they do not know if they have a reluctant Bush responder problem; even less do they know whether they have a reluctant Gore defector problem.

But to assume that the sample was random flies in the face of the evidence that weighting was required for visible characteristics. Even if selection was random, response wasn't, and never is.

Which is where the binomial theorem bites the dust.

[OnTheOtherHand]

108. or as it happened, a reluctant post-middle-aged responder problem

The interviewers reported response rates of 55.3% for ages 18-29; 55.6% for ages 30-59; 43.0% for ages 60+. But as you say, that's just what they could see.

Stuff like this is why Mitofsky and Lenski, in 2001, recommended raising the critical value for calls from 2.6 (about 300:1 nominal odds) to 4.0 (over 31,000:1) at poll closing time -- and, if that criterion wasn't met, not making a call until it could be made based on vote count data alone. Mitofsky and Lenski were nowhere near calling Ohio -- Kerry's apparent lead would have had to be probably three times as large. (And if it had been, they probably still wouldn't have called it, because the exit poll result would have been too far from prior expectations.) Of course that doesn't tell us what really happened in Ohio, it just tells us something about the exit polls.